Citation
Vehicle Dynamics Measurement System for Evaluating Older Driver Performance

Material Information

Title:
Vehicle Dynamics Measurement System for Evaluating Older Driver Performance
Creator:
DAVIS, ETHAN J. ( Author, Primary )
Copyright Date:
2008

Subjects

Subjects / Keywords:
Acceleration ( jstor )
Accelerometers ( jstor )
Aircraft maneuvers ( jstor )
Global positioning systems ( jstor )
Sensors ( jstor )
Signals ( jstor )
Simulations ( jstor )
Speed ( jstor )
Statistics ( jstor )
Steering wheels ( jstor )

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Ethan J. Davis. Permission granted to University of Florida to digitize and display this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Embargo Date:
7/24/2006
Resource Identifier:
496801924 ( OCLC )

Downloads

This item has the following downloads:

davis_e ( .pdf )

davis_e_Page_021.txt

davis_e_Page_059.txt

davis_e_Page_129.txt

davis_e_Page_004.txt

davis_e_Page_067.txt

davis_e_Page_027.txt

davis_e_Page_050.txt

davis_e_Page_101.txt

davis_e_Page_064.txt

davis_e_Page_042.txt

davis_e_Page_077.txt

davis_e_Page_134.txt

davis_e_Page_033.txt

davis_e_Page_118.txt

davis_e_Page_030.txt

davis_e_Page_128.txt

davis_e_Page_057.txt

davis_e_Page_062.txt

davis_e_Page_112.txt

davis_e_Page_044.txt

davis_e_Page_116.txt

davis_e_Page_003.txt

davis_e_Page_082.txt

davis_e_Page_106.txt

davis_e_Page_029.txt

davis_e_Page_130.txt

davis_e_Page_013.txt

davis_e_Page_052.txt

davis_e_Page_122.txt

davis_e_Page_034.txt

davis_e_Page_103.txt

davis_e_Page_023.txt

davis_e_Page_107.txt

davis_e_Page_015.txt

davis_e_Page_055.txt

davis_e_Page_097.txt

davis_e_Page_119.txt

davis_e_Page_010.txt

davis_e_Page_078.txt

davis_e_Page_035.txt

davis_e_Page_058.txt

davis_e_Page_115.txt

davis_e_Page_068.txt

davis_e_Page_096.txt

davis_e_Page_020.txt

davis_e_Page_007.txt

davis_e_Page_022.txt

davis_e_Page_012.txt

davis_e_Page_084.txt

davis_e_Page_113.txt

davis_e_Page_110.txt

davis_e_Page_031.txt

davis_e_Page_025.txt

davis_e_Page_019.txt

davis_e_Page_051.txt

davis_e_Page_092.txt

davis_e_Page_080.txt

davis_e_Page_109.txt

davis_e_Page_100.txt

davis_e_Page_091.txt

davis_e_Page_053.txt

davis_e_Page_094.txt

davis_e_Page_123.txt

davis_e_Page_041.txt

davis_e_Page_121.txt

davis_e_Page_093.txt

davis_e_Page_045.txt

davis_e_Page_102.txt

davis_e_Page_049.txt

davis_e_Page_105.txt

davis_e_Page_006.txt

davis_e_Page_008.txt

davis_e_Page_086.txt

davis_e_Page_056.txt

davis_e_Page_133.txt

davis_e_Page_002.txt

davis_e_Page_073.txt

davis_e_Page_104.txt

davis_e_Page_088.txt

davis_e_Page_072.txt

davis_e_Page_014.txt

davis_e_Page_085.txt

davis_e_Page_127.txt

davis_e_Page_032.txt

davis_e_Page_075.txt

davis_e_Page_132.txt

davis_e_Page_090.txt

davis_e_Page_009.txt

davis_e_Page_124.txt

davis_e_Page_028.txt

davis_e_Page_070.txt

davis_e_Page_005.txt

davis_e_Page_039.txt

davis_e_Page_040.txt

davis_e_Page_026.txt

davis_e_Page_074.txt

davis_e_Page_071.txt

davis_e_Page_076.txt

davis_e_Page_036.txt

davis_e_Page_017.txt

davis_e_Page_131.txt

davis_e_Page_125.txt

davis_e_Page_079.txt

davis_e_Page_038.txt

davis_e_Page_098.txt

davis_e_Page_016.txt

davis_e_Page_018.txt

davis_e_Page_061.txt

davis_e_Page_011.txt

davis_e_Page_001.txt

davis_e_Page_024.txt

davis_e_Page_081.txt

davis_e_Page_054.txt

davis_e_Page_069.txt

davis_e_Page_089.txt

davis_e_Page_108.txt

davis_e_Page_066.txt

davis_e_Page_060.txt

davis_e_Page_048.txt

davis_e_Page_099.txt

davis_e_Page_117.txt

davis_e_Page_095.txt

davis_e_Page_037.txt

davis_e_Page_083.txt

davis_e_Page_126.txt

davis_e_Page_120.txt

davis_e_Page_111.txt

davis_e_Page_043.txt

davis_e_Page_063.txt

davis_e_Page_047.txt

davis_e_pdf.txt

davis_e_Page_087.txt

davis_e_Page_114.txt

davis_e_Page_065.txt

davis_e_Page_046.txt


Full Text












VEHICLE DYNAMICS MEASUREMENT SYSTEM FOR EVALUATING OLDER
DRIVER PERFORMANCE
















By

ETHAN J. DAVIS


A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF ENGINEERING

UNIVERSITY OF FLORIDA


2006

































Copyright 2006

by

Ethan J. Davis
































This document is dedicated to my grandfather Robert 'Sport' Davis Sr.















ACKNOWLEDGMENTS

I would like to thank my advisor Dr. Niezrecki for assisting me with this research

and providing useful knowledge and guidance. I would also like to thank Dr. John

Schueller for helping me with many on-campus issues and Dr. Carl Crane for serving on

my supervisory committee. Furthermore, I would like to express gratitude to everyone at

the Occupational Therapy Department at the College of Public Health and Health

Professions. I would also like to thank Richard Phillips for helping me with parts of my

research and providing some helpful tips along the way.

Finally, I would like to thank my parents and the rest of my family for their

continued support.
















TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S ................................................................................................. iv

LIST OF TABLES .................. .................. ................... .......... .............. viii

LIST OF FIGURES ......... ....... .................... .......... ....... ............ ix

ABSTRACT .............. .......................................... xi

CHAPTER

1 IN TR O D U C T IO N ............................................................. .. ......... ...... .....

1.1 Background .............. ...................... .................. ...............1
1.2 O lder D river Statistics ................................................... ............................... 2
1.3 Purpose of FH W A R research ......................................................................... ....
1.4 O bjectiv es ................................................................. ............................... . 5

2 LITER A TU R E R EV IEW ............................................................... ...................... 9

2.1 Previous R research Studies ......................................................... ...............9
2 .2 D e v ic e s ............................................................................................................ 1 3
2 .2 .1 D riv e C a m .............................................................................................. 1 3
2 .2 .2 V ericom ........................................................................................... .... ...... 14
2 .2 .3 D ew etron .............................................................. 16

3 SYSTEM REQUIREM ENTS ................................... ....... ............... .... ...........18

3 .1 P e rfo rm a n c e ..................................................................................................... 1 8
3 .2 T ra n sd u c e rs ...................................................................................................... 1 9
3 .3 G P S .................................................................................................................. 2 0
3 .4 V id e o ................................................................................................................ 2 1
3.5 Ease of Use ................. ...... ..... .......... 21
3 .6 C o st ................................ ................................................................................. 2 1
3.7 F flexibility ......... ...... ............................................... ..........................22

4 SYSTEM DESIGN ................. ... .............. ........23

4.1 V vehicle D description ...........................................................23



v









4.2 Selection of Com ponents......................................................... .............. 26
4.2.1 D ata A acquisition System ........................................ ........................ 26
4 .2 .2 T ran sdu cers........ ................................................................ ...... ..... .. 2 8
4 .2 .3 M odu les .............................................................................. 3 1
4.2.4 Pow er Supplies .................................... ................ .. .. ................. 32
4.2.5 GPS ................ ......... ...... .......... 36
4 .2 .6 V id e o ................................................................3 6
4.2.7 Softw are ............. .............. ............ ....... ................... ......... 38
4 .3 In stallatio n ................................................................3 9
4 .4 C alib ratio n ................................................................4 9
4 .5 O operation ................................................................... 49

5 SYSTEM VERIFICATTON .............................................. ............... 51

5.1 Interpretation of D ata..................................................... 51
5.2 Analysis of Acceleration Test Maneuver .............................. ........ ........ 54
5.3 Analysis of Deceleration Test Maneuver ..................... ..............56
5.4 Analysis of Right Turn Test Maneuver ......................... ..............58
5.5 Analysis of Left Turn Test Maneuver ...................................... 60

6 FHWA RESEARCH ANALYSIS ....................................................................63

6.1 Description of main.m (Road Course) ............... ........... ..............................64
6.2 Description of stats.m (Road Course) ............... .................................. 67
6.3 Description of sort.m (Road Course) ............... .................................. 69
6.4 Description of sim.m and sim sort.m .............. ...................................... 69
6.5 D description of SPSS Analysis ..................................................... ...............70

7 RESULTS AND DISCU SSION ..................................... ............................. 73

7.1 FH W A R research R esults................................................. ................... ...... 73
7.2 Example Analysis of Turning Phase of FHWA Maneuver .............................73

8 C O N C L U S IO N ............................ .. .................................................................. 7 8

8.1 Future R research Studies ..................................................................... 78
8.2 Recommendations and Improvements............................................................ 79

APPENDIX

A INTERSECTION DIAGRAM S ........................................ ........................... 80

B COMPONENT TECHNICAL SPECIFICATIONS ..............................................92

C CU STOM PULLEY DRAW INGS ........................................................ ....... ........ 95

D M A T L A B C O D E S .......................................................................... .....................97









E EXAMPLE OF SPSS OUTPUT FILE FOR MAXIMUM YAW DURING THE
TURN PHASE OF M ANEUVER 1 .............................. ....................118

L IST O F R E F E R E N C E S ...................................................................... ..................... 120

BIOGRAPHICAL SKETCH ............................................................. ............... 122
















LIST OF TABLES


Table page

4-1 Technical Specifications for Signal Conditioning Analog Input Modules .............32

4-2 Transducer Output Ranges and Module Input Ranges...........................................32

5-1 Measurement Abbreviations and Respective Units for All Plots.............................53

7-1 Summary of Statistics for Turning Phase of FHWA Road Course Maneuver 2......74
















LIST OF FIGURES


Figure pge

1-1 Street Map of The FHWA Road Course in Gainseville, FL. ....................................4

1-2 Picture of NODRTC Simulator showing vehicle cab and projection screens...........7

2-1 Display Output from ARGOS Instrumented Vehicle (Rizzo et al., 292) ...............11

2-2 P picture of D riveC am ........................................................................ ........ ........... 14

2-3 Picture of Vericom VC3000 .............. ......................................... ..... ............ 15

2-4 Picture of Dewetron DEW E5000. ................................................................ ....... 16

4-1 Exterior View of 2005 Buick Century Sedan NODRTC Research Vehicle. ...........23

4-2 Picture of Left Foot Accelerator Pedal Adapter................... ..................................24

4-3 Picture of Hand Controls, Turn Signal Extender, and Steering Wheel
Receptacles ............... ..... ........... ........ .. ........ ........ ......... 25

4-4 Picture of Auxiliary Brake Pedal ................... ...............25

4-5 Wiring Diagram for Data Acquisition System ................. .............................. 27

4-6 Picture of National Instruments Signal Conditioning Carrier (SC-2345) and
M o d u le s .......................................................................... 2 7

4-7 Picture of Kistler Type 5210 Capacitive Accelerometer Power Supply.................33

4-8 Picture of PCB Model 480E09 Power Supply for ICP Laser Tachometer. ............34

4-9 Wiring Diagram of Angular Rate Sensor Power Supply. ......................................35

4-10 Picture of Equipment Installed in Trunk. ....................................... ............... 40

4-11 Trunk Diagram Displaying the Approximate Location of Each Component. .........40

4-12 Picture of Laser Tachometer, Mounting Bracket, and Drive Shaft.......................42

4-13 Picture of Installed Sting Potentiometer and Custom Pulley Assembly ................43









4-14 Picture of Lateral Accelerometer, Longitudinal Accelerometer, and Angular
Rate Sensor Installed Underneath the Front Passenger Seat...............................44

4-15 Picture of Miniature Video Camera and Installation Bracket...............................46

4-16 Wiring Diagram of All System Components. ................... .................47

4-17 Wiring Diagram Indicating the Location of All System Components in the
V vehicle. .............................................................................4 8

5-1 Example Plot of a Typical Left Turn. ........................................... ............... 54

5-2 Plot of a Typical Normal Acceleration Test Maneuver. ........................................55

5-3 Plot of a Typical Aggressive Acceleration Test Maneuver................................. 56

5-4 Plot of a Typical Normal Deceleration Test Maneuver. ........................................57

5-5 Plot of a Typical Aggressive Deceleration Test Maneuver............... ..................57

5-6 Plot of a Typical Normal Right Turn Test Maneuver....................... ...............58

5-7 Plot of a Typical Aggressive Right Turn Test Maneuver. .....................................59

5-8 Plot of a Typical Normal Left Turn Test Maneuver. .............................................61

5-9 Plot of a Typical Aggressive Left Turn Test Maneuver .......................................61

6-1 FHWA Road Course Generated Using GPS Coordinates with Intersections of
Interest H highlighted and Labeled. ........................................ ........................ 65

6-2 Example of Maneuver Plot Used to Verify That The Vehicle Dynamics Data and
GPS D ata Are Synchronized .................................. ............... ............... 66

6-3 Plot of Maneuver 4B Generated Using stats.m With Markers Indicating the Start
and End of The Turning Phase ........................................ ........................... 68

7-1 Picture of Right-Turn Channelization from FHWA Road Course Maneuver 2A.... 75















Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Engineering

VEHICLE DYNAMICS MEASUREMENT SYSTEM FOR EVALUATING OLDER
DRIVER PERFORMANCE


By

Ethan J. Davis

May 2006

Chair: Christopher Niezrecki
Cochair: John Schueller
Major Department: Mechanical and Aerospace Engineering

In 2003, The Occupational Therapy Department of The College of Public Health

and Health Professions at The University of Florida founded the National Older Driver

Research and Training Center (NODRTC). The NODRTC has received funding from the

Centers for Disease Control (CDC), Federal Highway Administration (FHWA), National

Highway and Traffic Safety Administration (NHTSA), and the Florida Department of

Transportation (FDOT). The NODRTC offers assessment of driving skills, referrals to

health professionals to improve driving skills, and counseling in the use of alternative

transportation for former drivers.

This thesis discusses the development and implementation of a system to measure

vehicle dynamics for the purpose of evaluating older driver performance. The system will

be used for several research studies, although this paper will only discuss its use in an

FHWA study. The goal of the study is to evaluate the effectiveness of intersection









improvements recommended by the FHWA. The recommended intersection

improvements are designed to make the intersections easier to negotiate, particularly for

older drivers.

The vehicle dynamics measurement system consists of a data acquisition system,

various sensors, and a GPS antenna. The system is capable of measuring lateral

acceleration, longitudinal acceleration, yaw (turning rate), speed, steering wheel position,

brake status, and turn signal status. There is also an event trigger that can be activated by

an evaluator to mark the data so that the event can be analyzed later. The GPS antenna is

used to determine the location of the vehicle while it is driving a pre-determined road

course. The data are extracted for the intersections of interest and an algorithm is used to

divide the intersection into three distinct phases (approach, turn, and recovery). Several

statistics are calculated for each variable and phase of the intersection. These statistics

were analyzed using SPSS statistical analysis software.

Overall the system was useful in acquiring vehicle dynamics at particular

intersections and inferring differences between old and young drivers. Furthermore, in

relation to the FHWA study, the results indicated that the recommended intersection

improvements benefited both old and young drivers and were more helpful to older

drivers.














CHAPTER 1
INTRODUCTION

This thesis will discuss the development of a vehicle dynamics acquisition system

to be used as an aid in driving research. The system will be installed in a 2005 Buick

Century Sedan to be used by the National Older Driving Research and Training Center

(NODRTC) to conduct research on older drivers. This chapter will discuss the

background and motivation for older driver research, specifically the research being

conducted by the NODRTC. The goals and objectives of the vehicle dynamics

acquisition system will also be presented.

1.1 Background

The Occupational Therapy Department of the College of Public Health and Health

Professions at The University of Florida founded the Seniors Institute on Transportation

and Communications in 2000. The name was later changed to the National Older Driver

Research and Training Center to reflect its natural focus after receiving funding in 2003

from the Centers for Disease Control and Prevention (CDC) and the Federal Highway

Administration (FHWA). The NODRTC later received additional funding from the

National Highway and Traffic Safety Administration (NHTSA) and the Florida

Department of Transportation (FDOT) and is currently conducting three research projects

for the CDC, NHTSA, and FHWA. The purpose of the CDC research is to determine the

effectiveness of particular screening tests in determining a person's driving abilities. The

goal is to identify which tests can accurately predict a driver to be unfit to drive and

prevent them from having to take a potentially dangerous driving test. The NHTSA









research is being performed to determine the validity and reliability of guidelines issued

by the American Medical Association (AMA) for diagnosing driving ability, specifically

The Physician's Guide to Assessing and Counseling Older Drivers. The FHWA research

evaluates older driver performance on roadway design features identified by the FHWA

as possibly problematic for older drivers. The FHWA has issued guidelines for highway

designs that are believed to increase the safe driving ability of older drivers. This research

examines differences in driving performance between road designs that either meet or do

not meet these guidelines.

The NODRTC has two vehicles dedicated to driving research and evaluations. The

design and implementation of the system described in this thesis may eventually be used

for all of the research projects previously mentioned. Initially it will be used for the

FHWA research and for that reason it will be the focus of the discussion. The system

itself will be the same for the different studies. However, the specific goals for acquiring

and processing the data will be unique. From this point on, all discussion will be relevant

to the FHWA research, particularly the data processing and analysis.

1.2 Older Driver Statistics

The emphasis for evaluating older driver's driving abilities is motivated by the

growing number of senior citizens driving, their frequency of collisions, and the severity

of their crashes. In 2003, there were 35.9 million people in United States age 65 or older.

That number is expected to double by the year 2030. When vehicle fatality rates are

compared on the basis of miles traveled annually, the fatality rate for seniors is four times

that of the 30-59 year old age group (DOT HS 809 328). An interesting note is that most

of the older driver fatalities in 2000 occurred during the daytime (81%), on weekdays

(71%), and involved another vehicle (76%) (DOT HS 809 328). All of these statistics









suggest that older drivers are some of the most dangerous drivers on the road.

Furthermore, with an inevitable increase in the number of older drivers from the baby

boom generation, the number of collisions and fatalities as a result of older drivers is

likely to increase.

1.3 Purpose of FHWA Research

As previously mentioned, the FHWA research project is focused on assessing older

driver performance on roadway conditions identified by the FHWA as potentially

problematic for older drivers. The FHWA has proposed certain roadway improvements

that are intended to help all drivers negotiate an intersection. It is believed that these

improvements will be most beneficial to older drivers. Two examples of these roadway

improvements are an offset turning lane and a four foot wide receiving lane. The offset

turning lane is the left-most lane when approaching an intersection. Instead of facing the

oncoming turning lane, which is the standard for most intersections, the offset turning

lane is positioned four or five feet further left. This results in the driver having an

improved view of oncoming traffic, thus making it easier and safer to complete the turn.

The four foot wide receiving lane is also intended to make navigation of an intersection

easier. The four foot wide receiving lane is an extension on the right side of the roadway

after completing a left turn at an intersection. The widened receiving lane allows the

driver to make a wider turn and avoid having to turn sharply and possibly hit the median

in the center of the road.

The NODRTC designed a road course in Gainesville, Florida, to evaluate the

effectiveness of these improved roadway safety features. The road course incorporates

ten intersections. Five of the intersections feature different recommended roadway

improvements and the other five are their similar counterparts without the improvements.







4


They are all signalized intersections and all but one are left turn maneuvers. A street map

of the FHWA road course showing the route is provided in Figure 1-1.







S. State Park.
I" 2
'- i ,






I 2.

S. a





-' .]'


f S, Gale Park



Laee
.,, -

I










Figure 1. Street Map of The FHWA Road Course in Gainseville, FL.






A Driving Rehabilitation Specialist accompanies the test subject as they drive the
road course. The Driving Rehabilitation Specialist, also referred to as the evaluator,








assesses the driver's performance during the designated intersections. The evaluator uses
a scoring sheet that lists all of the possible driving errors for each intersection. As the
S"" i '--" -









driver navigates the course, the evaluator pays careful attention to each intersection and

scores driving errors including lane maintenance, forward vehicle gap acceptance, and

overall stability. The FHWA research calls for at least 100 drivers age twenty-five or

older to be evaluated. There are two hypotheses being simultaneously tested in the

FHWA study. The first is that all drivers will benefit from the improved safety features

and will have fewer driving errors on these particular intersections. The second

hypothesis is that older drivers will benefit from the new safety features even more than

the rest of the population. The older drivers are expected to have more driving errors per

person for the unimproved intersections, but perform similar to the younger drivers for

the improved intersections. It is certainly possible that any age driver may exhibit poor

driving performance for both the improved and unimproved intersections. This is why a

statistical analysis will be performed on the behavioral data to detect any underlying

trends.

1.4 Objectives

In addition to the subjective evaluator scoring, the NODRTC wanted an objective

way of measuring driver performance. The decision was made to design a system capable

of measuring and recording vehicle dynamics such as acceleration and yaw. The system

needed to be affordable and easy for an evaluator to operate. The desired measurements

were longitudinal acceleration, lateral acceleration, yaw, speed, brake status (on/off), and

steering wheel position. Longitude and Latitude provided by the Global Positioning

System (GPS) were also needed to determine when the vehicle was at a particular

intersection. The vehicle dynamics measurements will ultimately be combined with the

behavioral data to determine which kinematic measurements are the best indicators of









driving performance. This will be performed by a NODRTC research assistant once all of

the subjects have been evaluated.

Due to uncontrollable factors such as traffic and weather, each FHWA participant

will encounter different driving scenarios. This will certainly make it more difficult to

evaluate driver performance based on the vehicle dynamics. For example, if a subject

approached an intersection to make a left turn and had a green left-turn arrow, they would

most likely slow down to make the turn and then accelerate up to a safe driving speed. On

the other hand, if another subject approached the same intersection but had a red light,

they would have to stop and wait for a green light to complete the turn. The longitudinal

acceleration profiles for the two subjects would be drastically different. The first subject

would most likely decelerate before the turn, while the other subject has to accelerate

after being stopped at the red light. Therefore, drivers will need to be compared based on

the driving conditions that they encounter. In order to evaluate all of the drivers for the

FHWA study on an equal playing field, the NODRTC purchased a full-scale driving

simulator. The simulator features a vehicle cab and three large projection screens for 1800

viewing and a realistic driving experience. A picture of the simulator is provided in

Figure 1-2.

The ten intersections from the road course have been recreated for the simulator

using the driving scenario programming language. The simulator intersections were

designed to be as similar to their road course counterparts as was possible using the

scenario programming language. The simulator has some limitations including only being

able to produce perpendicular intersections. Since one of the road course intersections is

not perpendicular, it could not be recreated for the simulator. The advantage of evaluating









drivers with the simulator is that they can all be scored for the exact same driving

conditions. The simulator is capable of calculating all of the vehicle dynamics that are

being measured on the road course. Since all drivers will encounter the same scenarios,

the simulator should be helpful in determining which measures are good indicators of

driving performance.


Figure 1-2. Picture of NODRTC Simulator showing vehicle cab and projection screens.

The rest of this paper will focus on the design, selection, installation, and data

processing of the road course vehicle dynamic measuring system. A literature review of

similar systems and driving research using such systems is provided in Chapter 2. The

specific requirements for the system will be discussed in detail in Chapter 3. The overall

design of the system and the selection and installation of components will be discussed in


,.T.






8


Chapter 4. System verification and sample driving maneuvers will be analyzed in Chapter

5. Chapter 6 will review the processing and analysis of the data as it relates to the FHWA

study. The results and discussion will be presented in Chapter 7. Finally, a conclusion is

provided in Chapter 8.














CHAPTER 2
LITERATURE REVIEW

A literature review was conducted to gather information for two purposes. The first

was to investigate other research studies which in some way used vehicle dynamics

and/or a quantitative assessment to evaluate driving performance. The second was to

evaluate off-the-shelf or "turn-key" devices which can be used to measure and record

vehicle dynamics. It should be noted that the NODRTC conducted an extensive literature

review of more than 2,500 articles related to elderly driving, driving rehabilitation, and

intersection roadway design. The literature review presented in this chapter is solely

concerned with instrumented vehicles, devices for acquiring vehicle dynamics, and

relevant research studies.

2.1 Previous Research Studies

Wood and Worringham (2005) used a quantitative assessment to evaluate driving

performance of drivers with Parkinson's disease. An accredited professional driving

instructor and a trained occupational therapist accompanied each driver as they navigated

a 19.4 kilometer open road course. The occupational therapist "scored seven aspects of

driving performance at 147 locations along the route (general observation, observation of

blind spots, indication (signaling), braking/acceleration, lane position, gap selection, and

approach" (177). The occurrence and type of error was recorded for each location. Also,

each of the 147 locations "was allocated to one of nine categories, including roundabouts,

merging, pulling in/pulling out, traffic light and non-traffic light controlled intersections,

reversing and parking, one way and two way traffic, and lane changing" (177). This









allowed researchers to determine what types of errors were occurring most frequently and

in which type of driving situation. The date was used to give each driver a driver safety

rating. The study was used to correlate a driver's driver safety rating with the number of

years since the diagnosis of Parkinson's disease. Although this research did not make use

of an instrumented vehicle, it showed how an objective quantitative assessment could be

used to evaluate driving performance.

Knodler et al. (2005) used a quantitative assessment and a "full-scale dynamic

driving simulator as a tool for evaluating driver comprehension by placing drivers in a

fully interactive scenario just as if they were actually driving" (270-271). The goal of the

study was to determine which type of protected/permissive left-turn (PPLT) signal

display was most effective. Again, this study did not make use of an instrumented vehicle

however it used a full scale simulator to objectively measure driver behavior.

Another simulator study was conducted by McGehee et al. (2001) to investigate

how long it takes for a participant to adapt to the simulator. McGehee et al. objectively

measured a participants adaptation by measuring steering wheel reversals. "A steering

wheel reversal is operationally defined as a deflection of the steering wheel away from a

central or neutral position (i.e., where the wheels are completely straight) followed by a

reversal of the steering wheel back to the neutral position" (2). In order to quantify

steering wheel reversals a threshold of 6 degrees from neutral position was chosen.

Steering wheel reversals were counted during a 25-minute drive and were used to

determine how quickly a driver could adapt to the simulator. The study showed that on

average "older drivers need about three minutes to adapt and get the "feel" of the









simulator. Before this time driving behavior in the simulator may not be representative of

actual driving performance" (4).

Rizzo et al. (2002) conducted a study using an instrumented vehicle because "road

tests were designed to ensure that novice drivers know and can apply the rules of the

road, and not to test older, experienced drivers who may be impaired" (291). Rizzo et al.

also states "an instrumented vehicle permits quantitative assessments of driver

performance in the field, under actual road conditions. These measurements are not

subject to the type of human bias that affects inter-rater reliability on a standard road test"

(291-292). The researchers developed an instrumented vehicle called ARGOS (the

Automobile for Research in Ergonomics and Safety). "ARGOS was designed to examine

objective indices of driving performance in normal and potentially unfit drivers and

consists of a mid-seized vehicle with extensive instrumentation and sensors hidden within

its infrastructure" (292). ARGOS also utilized video cameras and an example of the

display output from ARGOS is shown in Figure 2-1.


LTACC -0.01 SPEEDI2
FRAME 3097 BRAKE 11


Figure 2-1. Display Output from ARGOS Instrumented Vehicle (Rizzo et al., 292)


~""e~
- -









The four video images were multiplexed onto a single display. The values for the various

measurements were superimposed over the video display. Uc et al. (2004) used ARGOS

to evaluate route-following and safety errors in drivers with early Alzheimer disease and

also provided a more detailed description of ARGOS. "Experimental performance data

(e.g.,, steering wheel position, normalized acceleration and vehicle speed) were digitized

at 10 Hz and reduced to mean SD (standard deviation), or count. Driver's lane tracking

and visual scanning activity of the environment were recorded by videotape at 30

frames/s using miniature lipstick-sized cameras mounted unobtrusively within the

vehicle" (833). It is evident from Figure 2-1 that ARGOS has forward, rear, driver's

face, and driver's hands cameras. Furthermore, ARGOS was capable of measuring

accelerator (gas) pedal and brake pedal depression (measured as a percent of fully

depressed). ARGOS was used successfully to quantitatively assess drivers with

Alzheimer disease while performing a specific driving task.

Dingus and Hulse (1997) used an instrumented vehicle to determine the effects of

age and experience on navigation while driving. "The system consisted of video cameras

to record pertinent event and eye movement data, an experimenter control panel to record

incident classification data, and sensors for the detection of variations in driving

performance and behavior" (185). The system utilized four video cameras with nearly

identical views as those used with ARGOS. "The system provided the capability to store

a line of numerical data every 0.1 seconds" (185). It utilized accelerometers to measure

lateral and longitudinal acceleration forces. The system also measured steering wheel

position, vehicle speed, and brake pedal status (on/off). The research focused on using

lateral acceleration measurements to indicate driver awareness. It also made use of the









video cameras to time the driver's glancing eye movements. Although the study was

mainly concerned with driver navigation and awareness rather than driver performance, it

was able to use an instrumented vehicle to objectively quantify driver behavior.

Boyce and Geller (2002) used a vehicle with concealed video cameras and

transducers to evaluate driver behavior, specifically younger males. The "computer

generated dependent measures, in concert with real-time covert video recording of the

participants' driving, allowed for unprecedented opportunities to perform a behavior

analysis of driving performance in the context of normal driving. Thus, problems

associated with truthfulness of self-report, infrequency of occurrence, and error variation

due to in-vehicle observers were minimized or avoided altogether" (52). In addition to

video recording, the vehicle also measured and recorded driver safety-belt use, number of

times a turn signal was used, "vehicle velocity including average speed, velocity changes,

and velocity variance, and following distance measured in meters" (53). Again, the video

camera configuration was similar to that found in ARGOS. Although not directly related

to driver performance, this study made use of an instrumented vehicle to measure driver

behavior in the absence of a driving evaluator.

2.2 Devices

This section will offer a brief review of several commercially available products

that are used to measure vehicle dynamics. These devices are called turn-key devices

because they can be installed and ready to use right out of the box. In general, they are

designed to serve a single purpose and usually not expandable or configurable.

2.2.1 DriveCam

The DriveCam is a device that is primarily used for vehicle fleet tracking and

maintenance. It is installed behind the central rear view mirror in a vehicle. It features









two video cameras to record forward-facing and rear-facing views. In addition, it has a

longitudinal and a lateral accelerometer. A picture of the DriveCam is provided in Figure

2-2.


















Figure 2-2. Picture of DriveCam.

The DriveCam uses removable digital media and as a result can only store small

amounts of data (about 20 seconds). It records continuously and in the case of an event

(e.g., hard turn or braking), it will store 10 seconds before and 10 seconds after the event.

The measurements from the accelerometers are displayed at the bottom of the screen

during video playback. The accelerometers are only accurate to a tenth of a g. The device

can also be triggered by the user to save a specific event. Due to the limited number of

measurements and lack of expandability, in addition to the short recording time, the

DriveCam was not a suitable solution for the FHWA research.

2.2.2 Vericom

Vericom Computers Inc. manufactures the VC3000 which is an on board vehicle

dynamometer. It features a two axis accelerometer which is used to measure longitudinal

and lateral acceleration. It integrates the longitudinal acceleration to calculate speed and









distance traveled. It is a self contained unit and is usually installed with suction cups to

the inside of the passenger side windshield. The VC3000 can accept up to 6 additional

sensor inputs. These sensors are sold by Vericom and only a limited selection is

available. The data collected with the VC3000 can be downloaded to a computer and

analyzed using the provided software. Figure 2-3 shows a picture of the Vericom

VC3000.





















Figure 2-3. Picture of Vericom VC3000.

The VC3000 is primarily used for vehicle performance tests (e.g., horsepower,

braking) and traffic accident investigation and reconstruction. These tests are usually of

short duration and consequently the VC3000 was designed with only enough flash

memory to store approximately 21.8 minutes of data. This was the main reason that the

VC3000 could not be used for the FHWA research. Furthermore, it is not completely

expandable and the software has limited capabilities.
expandable and the software has limited capabilities.









2.2.3 Dewetron

Dewetron is a worldwide manufacturer of portable data acquisition systems. They

offer several different products depending on the requirements of the application. Their

products are essentially personal computers combined with data acquisition hardware and

Dewetron's own proprietary software. The units are capable of measuring nearly any type

of signal including video and GPS. Dewetron offers desktop as well as portable products

including some that are battery powered. Figure 2-4 shows a picture of one of the

Dewetron units.

,R






-i












Figure 2-4. Picture of Dewetron DEWE5000.

The Dewetron systems are very sophisticated and the software is specifically

designed for evaluating vehicle performance. However, in the context of the FHWA

research, the Dewetron system would only be used as a data recorder. The software used

to analyze the data from the FHWA road course would still have to be developed. As a






17


result, the Dewetron system would be too expensive to use especially considering that its

full functionality is not being utilized.














CHAPTER 3
SYSTEM REQUIREMENTS

After reviewing several off-the-shelf products with their own advantages and

disadvantages, a decision was made to design a system from scratch that satisfies all of

the requirements. The system must be capable of measuring accelerations, speed, yaw,

steering wheel position, longitude, latitude, and brake and turn signal status. Additionally,

the system should be affordable, easy to operate, and flexible. Designing a customized

system allows it to be configurable and used for a wide variety of applications. This

chapter will discuss the specific goals and performance requirements of the system as

well as additional considerations including cost, ease of use, and flexibility.

3.1 Performance

The system should offer good performance in the form of accuracy and high sampling

rate. This system will likely be used for future studies other than the FHWA research.

Therefore, it is necessary that the system offers a high level of performance such that

future upgrades will not be necessary. The system should have a resolution of at least 16-

bits to provide accurate sampling of each channel. A 16-bit system would allow a voltage

resolution of 0.0001526 volts for an input range of 10 volts. In addition to accuracy,

sampling rate is another important performance specification. A digital data acquisition

system can only sample an analog signal at discrete times. Therefore, a high sampling

rate is needed to accurately reproduce the original analog signal. If frequency

measurements are desired, the sampling frequency must be at least twice the highest

frequency component of the analog signal according to the Nyquist Theorem. This is to









prevent aliasing, which occurs when the sampled signal appears to have a different

frequency than the original analog signal. Frequency measurements, for example

frequency of steering wheel reversals, are not being considered for the FHWA study.

However, it will still be necessary to accurately recreate the original signal. Furthermore,

frequency measurements (e.g., steering wheel input) may be utilized in future NODRTC

research. It was decided that the system should be capable of sampling each channel at a

minimum of 10,000 samples per second (10k Hz). This will ensure accurate

reconstruction of the analog signal and allow for frequency measurements up to 5000 Hz.

This is certainly capable of measuring any input physically generated by a human. Also,

it is doubtful that any of the vehicle dynamics will have frequency components of this

magnitude.

3.2 Transducers

The accelerometers used for measuring longitudinal and lateral acceleration need to

be able to accurately measure at low frequencies (i.e., less than 10 Hz). Measurement of

the high frequency vehicle vibration is not important to this study. The changes in

acceleration will be gradual and the maximum acceleration in any direction under normal

driving conditions is not expected to exceed 1 g.

The transducer that is used to measure the steering wheel position should be non-

intrusive. It should be able to accurately provide the position of the steering wheel at any

instant. The sensor should not add any extra resistance to the steering wheel that may

affect the driver's ability to turn the wheel.

As with the accelerometers, the transducer that measures yaw should be able to

measure accurately at low frequencies. Under normal driving conditions, the maximum

yaw is not expected to exceed 900/sec. This would be the equivalent of completing a 900









turn in one second, which is unrealistic for everyday driving. The yaw sensor should be

able to determine the direction and rate that the vehicle is turning.

The transducer used to measure vehicle speed should be non-intrusive. Velocity

can be derived from the longitudinal acceleration, but a more direct measurement would

be preferable. It may be possible to acquire vehicle speed via the car's on board

diagnostic system.

The data acquisition system should be able to determine if the driver is using the

brake and the turn signals. Additionally, the evaluators requested that an event trigger be

installed in the vehicle cab. The event trigger will be used as a way to mark the data in

the case of an adverse event for later review.

3.3 GPS

A global positioning system receiver and antenna is needed to acquire the vehicle's

location while driving on the road course. The receiver should be WAAS capable and

easily integrated into the data acquisition system. The Wide Area Augmentation System

(WAAS) consists of 25 ground reference stations and two geostationary satellites orbiting

above North America. The ground reference stations monitor the GPS satellites and

correct for errors caused by satellite position error, clock drift, and atmospheric effects

(Garmin Ltd.). The corrected signals are then relayed to the geostationary satellites. The

correction message can be received from the geostationary satellites using a standard

WAAS-capable GPS receiver. A GPS receiver using WAAS has an accuracy of less than

3 meters 95 percent of the time. Differential GPS (DGPS) uses a local reference receiver,

whose location is known exactly, to calculate the error in the GPS signals. The error

signals are then broadcast, usually via radio, to the other GPS receivers. DGPS requires a









reference receiver and radio transmitters and is consequently more costly than a WAAS-

capable receiver. As a result, DGPS will not be considered for this project.

3.4 Video

The video system should consist of a forward view, rear view, and a view of the

driver. The driver view will be used after a driving assessment to evaluate the driver's

eye movements. All of the views should be recorded simultaneously to a single digital

medium. Digital video will provide distortion-free frame by frame playback. The media

needs to be long enough so that the entire road course will fit on a single recording.

3.5 Ease of Use

The data acquisition system will be primarily operated by Occupational Therapists

and therefore the system should be user friendly and able to accommodate users with

minimal technical experience. The end user should not have to calibrate or configure the

system in any way. Consideration will be taken in the design of the system such that the

user can operate it after receiving minimal training.

3.6 Cost

The NODRTC has received substantial funding from the FHWA. In addition to the

data acquisition system, the budget must be used for purchasing the road course vehicle,

driving simulator equipment, and vehicle cab for the driving simulator. Therefore the

costs for the vehicle dynamics acquisition system should be minimized if possible.

Regardless of the data acquisition system used, the transducers and sensors will cost the

same. The system should be expandable so that the option to use additional sensors for

the FHWA or other research is available.






22


3.7 Flexibility

This system will likely be used for other older driver research at the conclusion of

the FHWA study. The system must be flexible and adaptable to other areas of research.

For example, the FHWA research is only concerned with 10 specific intersections

whereas other research may be interested in driver performance on an entire course.

Furthermore, different measurements, possibly physiological, may be desired for future

research.














CHAPTER 4
SYSTEM DESIGN

This chapter will discuss the selection, installation, and operation of the entire

system including video. The selection of components including the data acquisition

system, transducers, GPS, and video equipment will be discussed in detail. A description

of the vehicle and its driving aids will also be presented.

4.1 Vehicle Description

The system was installed in a 2005 Buick Century Sedan. A picture of the vehicle

is provided in Figure 4-1.


















Figure 4-1. Exterior View of 2005 Buick Century Sedan NODRTC Research Vehicle.

The vehicle is equipped with adaptive driving equipment for handicapped drivers.

There is a left foot accelerator pedal adapter that mounts to the floor and replaces the

standard accelerator pedal. The adapter is secured to a bracket that is permanently

attached to the floor using locking pins. This adaptation is for drivers who have limited or









no use of their right leg. The adapter features a metal plate to prevent the driver from

accidentally pressing the original accelerator with their right leg. Figure 4-2 shows a

picture of the left foot accelerator adapter.


























Figure 4-2. Picture of Left Foot Accelerator Pedal Adapter.

For driver's who have limited use of both legs, there are hand controls. The hand

control is a lever on the left side of the steering wheel that is mechanically linked to the

brake and accelerator pedals. The driver rotates the lever to accelerate and pushes it

forward to brake. Since the driver must use their left hand to control the accelerator and

brake, there is a turn signal extender that allows the driver to activate the turn signal with

their right hand. The car also features three receptacles on the steering wheel for spinner

knobs. The knobs enable a driver suffering from arthritis to turn the steering wheel easier.

A picture of the hand controls and steering wheel receptacles is shown in Figure 4-3.































Figure 4-3. Picture of Hand Controls, Turn Signal Extender, and Steering Wheel
Receptacles.

In addition to the adaptive driving equipment, the car is also equipped with an auxiliary

brake pedal for the evaluator. The auxiliary brake pedal is shown in Figure 4-4.


Figure 4-4. Picture of Auxiliary Brake Pedal.









The auxiliary brake pedal is mechanically linked to the standard brake pedal using a steel

cable and pulleys. The auxiliary brake is used by the evaluator if a driver exhibits

reckless driving behavior. Also, the car was purchased with a bench seat which allows the

evaluator to grab the steering wheel and take control of the vehicle if necessary.

4.2 Selection of Components

4.2.1 Data Acquisition System

The data acquisition system is the centerpiece of the design. All of the sensors are

connected to it and it is used to record all data. It is also the user interface for the entire

system. The system was installed in a car and must be mobile, so a desktop computer

would not be practical. The decision was made to use a powerful laptop mated with a

data acquisition card. The laptop chosen is a Dell Inspiron with a Pentium 4 2.4 GHz

processor and 512 MB of RAM. The data acquisition card is a National Instruments

DAQCard-6036E with PCMCIA connectivity. The card is 16-bit and is capable of

200,000 samples per second. This allows sampling rates of 10,000 samples per second for

20 channels. The card is connected to a Signal Conditioning Carrier (SCC) via a 68-pin

cable. The SCC is the link between the sensors and the data acquisition card. Each SCC

can hold several signal conditioning modules, which are customized inputs depending on

the type of sensors used. The modules allow connection of different transducers (e.g.,

accelerometers, thermocouples, and strain gauges) to a single measuring device. The SCC

most appropriate for this application is the SC-2345. The SC-2345 can hold up to 20

modules (16 inputs, 2 outputs, 2 counters) and has 8 digital I/O lines built in. There are

three options for powering the SC-2345: 5 volt DC power from a DAQ device, AC

power, or a 7 to 42 volt DC power supply. AC power is required to power the laptop, so it

will also be used to power the SC-2345. In addition to selecting the right module for each









sensor input, the physical connection can also be specified. There are a wide variety of

connections including BNC, 9-pin D-sub, SMB, MIL-Spec, and LEMO. These

connections are referred to as panelettes and any combination may be used. For this

application, all of the panelettes are BNC and have two inputs per panelette. The

panelettes have lead wires which are connected to the screw terminals of each module.

This allows each sensor to be connected using its native connector. A simple wiring

diagram is shown in Figure 4-5 and a picture of the SCC is provided in Figure 4-6.


Figure 4-5. Wiring Diagram for Data Acquisition System.


Figure 4-6. Picture of National Instruments Signal Conditioning Carrier (SC-2345) and
Modules.

The selection of modules will be discussed following the description of the sensors. The

sensors must be selected first so that the proper type and input range of each module can

be chosen.


Transducer Inputs









4.2.2 Transducers

The section will discuss the selection of each transducer. The basis for selection

and relevant specifications will be provided.

Accelerometers. It was determined that the accelerometers used to measure

longitudinal and lateral acceleration need to be capable of measuring at extremely low

frequencies. The decision was made to use a pair of static accelerometers manufactured

by Kistler. These accelerometers can measure accelerations in the range of 0 to 300

Hertz. The accelerometers utilize a variable capacitance sensing element. A small inertial

mass is attached to a flexure element cantilevered between two plates. The mass deflects

under acceleration resulting in a change in the capacitance between the two plates.

Internal circuitry provides AC excitation and synchronous amplitude demodulation

resulting in a voltage output that is proportional to the applied acceleration. There are two

models available (8310A2 and 8310A10) with ranges of either 2 or 10 g's. The 10 g

model was chosen to measure longitudinal acceleration while the 2 g model was chosen

for lateral acceleration. The 2 g model is used to measure lateral acceleration because

even a high-end sports car cannot exceed Ig laterally on a skidpad (Road & Track). By

choosing the accelerometer with the smaller range, the lateral acceleration can be

measured more precisely. The 10 g model was chosen to measure longitudinal

acceleration. Under normal driving conditions the longitudinal acceleration is not

expected to exceed 2 g's in either direction. However, it may be possible to exceed 2 g's

under heavy breaking or in a collision. Therefore the 10 g accelerometer is used to

measure longitudinal acceleration to prevent saturation of the signal at 2 g's. Technical

specifications for the accelerometers and all other transducers are provided in Appendix









B. Installation and power requirements for the accelerometers will be discussed later in

the chapter.

Angular rate sensor (Yaw). An angular rate sensor was chosen to measure the

vehicle yaw (i.e., rate of rotation). The angular rate sensor is manufactured by BEI

Technologies Inc., and is called the GyroChip (AQRS-00075-111). It is a compact,

rugged micromachined vibrating quartz gyroscope, designed specifically for demanding

automotive and commercial applications (BEI 964172 Rev. A). The GyroChip uses a

double-ended quart tuning fork made from monocrystalline piezoelectric quartz. As a

result of the Coriolis effect, there is a voltage output directly proportional to the rate of

rotation. The use of a single piezoelectric crystal makes it less susceptible to variations in

temperature and extends its life. The maximum rate of rotation the GyroChip is capable

of measuring is 75 degrees per second. It is expected that this rate would not be exceeded

under normal driving conditions. If the car were to exceed this rate and sustain it, that

would mean it is capable of completing a 90 degree turn in just over a second. This is

certainly not realistic, however a strong jerk on the steering wheel may cause the car to

approach the maximum rate. Furthermore, the GyroChip is designed especially for use

with automotive stability control systems so it is an appropriate choice for measuring

vehicle yaw.

String potentiometer (steering wheel position). Measuring the position of the

steering wheel accurately and in a non-obtrusive manner proved to be difficult. An

absolute rotary encoder was first considered, but the decision was made to use a string

potentiometer and custom made pulley. It is not possible to mount a rotary encoder on the

steering shaft so it would have to be connected using a pulley or gear. Knowing that a









pulley or gear would be needed to connect a sensor to the shaft, a string potentiometer

was chosen instead. Most absolute rotary encoders transmit a digital word to the data

acquisition system. It is much easier to use a string potentiometer because it outputs a

voltage directly proportional to displacement. When attached to a pulley on the steering

shaft, it could accurately measure the position of the steering wheel. A PT1DC Cable-

Extension Position Transducer (PT1DC-50-BK-Z10-MC4-SG) produced by Celesco was

selected. Several aspects of the PT1DC are customizable including cable length, cable

exit location, output voltage, and connection type. The PT1DC features internal power

conditioning so it does not require a regulated power supply. The cable tension is only 5

ounces which will apply a negligible amount of torque (less than 0.1 ft-lb.) to the steering

shaft. This small amount of torque will have negligible effect on the position of the

steering wheel or the effort needed to rotate it in either direction. The maximum available

range of 50 inches was selected in case the string potentiometer could not be mounted

close to the steering shaft. If absolutely necessary, several pulleys could be used to route

the cable from the sensor to the pulley on the steering shaft. The design of the custom

made pulley for the steering shaft and the installation of the string potentiometer will be

discussed later in the chapter.

Laser tachometer (Speed). As with steering wheel position, measuring speed in

a non-obtrusive manner proved to be a challenge. Obtaining vehicle speed from the On-

Board Diagnostic system was investigated but proved to be too difficult to integrate with

the data acquisition system. The LaserTach ICP tachometer from The Modal Shop Inc.

was chosen to measure vehicle speed. The LaserTach emits a laser and has a sensor to

detect when it is reflected. A retroreflective target is placed on the rotating shaft to reflect









the laser. The LaserTach produces a voltage pulse each time the laser is reflected. The

frequency of the pulse signal is the rotational frequency of the shaft. Integrated Charge

Pre-Amplifier (ICP) sensors have a built-in microelectronic amplifier which results in a

low impedance coaxial system. An external ICP power supply or a data acquisition

system capable of powering ICP sensors must be used to provide power to the sensor.

Event trigger, brake, and turn signals. The event trigger, brake, and turn

signals are essentially identical. They are simply coaxial cables connected to each signal

individually. The coaxial cables are wired in parallel with the brake and turn signal lights.

The other end of each coaxial cable is connected to the data acquisition system. The event

trigger is slightly different in that it must be supplied a voltage. The event trigger is a

simple push-button with one terminal connected to +12 volts from the vehicle's electrical

system and the other terminal wired to the data acquisition system. The event trigger

input to the data acquisition system is coupled with a ground wire to determine when the

trigger is pushed. All of these signals will generate approximately 12 volts when

activated. A simple logic statement can be implemented in the software such that if the

voltage is above a certain threshold the signal will be considered activated.

4.2.3 Modules

As previously mentioned, the input modules for the SCC can be selected to match

the output of each sensor. Once all of the transducers were selected, the appropriate

modules could be chosen. All of the sensors chosen will require modules with analog

inputs. Each analog input module contains two individual inputs and are available with

different input voltage ranges. A total of five modules will be needed to accommodate all

of the inputs. It was determined that (2) 5 volt, (1) 10 volt, and (2) 20 volt modules

would be needed. The specifications for each module are provided in Table 4-1.









Table 4-1. Technical Specifications for Signal Conditioning Analog Input Modules.
Model Quantity Channels Input Range Bandwidth
SCC-AI02 2 2 + 20 volts 10 kHz
SCC-AI03 1 2 + 10 volts 10 kHz
SCC-AI04 2 2 + 5 volts 10 kHz
Each channel of each analog input module contains an instrumentation amplifier, a

lowpass filter, and a potentiometer for calibration. Table 4-1 displays the output range for

each transducer and the input range for the selected module.

Table 4-2. Transducer Output Ranges and Module Input Ranges.
Transducer Output Range (V) Module Input Range (V)
Longitudinal Accelerometer 0 -2 SCC-AI04 5
Lateral Accelerometer 0 -2 SCC-AI04 5
Angular Rate Sensor 0.25 4.75 SCC-AI04 5
String Potentiometer 0-10 SCC-AI03 10
Laser Tachometer 0- 1 SCC-AI04 5
Brake 0 12 SCC-AI02 20
Left Turn Signal 0 12 SCC-AI02 20
Right Turn Signal 0 12 SCC-AI02 20
Event Trigger 0 12 SCC-AI02 20
All of the modules will be connected to panelettes with BNC connections. BNC offers a

quick and reliable connection and is also the native form of connectivity for all of the

sensors.

4.2.4 Power Supplies

The accelerometers, angular rate sensor, and laser tachometer all require

specialized power supplies. The string potentiometer power supply does not need to be

regulated, nor does the supply voltage for the event trigger. The brake and turn signal

connections will be powered by the voltage provided to the respective lights.

The power supplies for the accelerometers are also manufactured by Kistler and run

off of a single 9-volt battery. The power supplies are specifically designed for capacitive

accelerometers. Each power supply is connected to an accelerometer via a 4-pin

Microtech connector. The output of each power supply uses a BNC connector and is









easily connected to the data acquisition system. Each power supply provides the

appropriate voltage and current to each accelerometer. They feature internal jumpers to

adjust the gain and a potentiometer to adjust the output voltage bias. The model is Type

5210 and a picture is provided in Figure 4-7.


Figure 4-7. Picture of Kistler Type 5210 Capacitive Accelerometer Power Supply.

The laser tachometer is an ICP sensor so a standard off-the-shelf ICP power supply

may be used. The selected power supply is a Model 480E09 manufactured by PCB. The

power supply uses BNC connections for both the input and output. It features an

adjustable gain and provides the proper voltage and current for the laser tachometer. The

power supply operates off of three 9-volt batteries providing approximately 50 hours of

battery life. A picture of the power supply is provided in Figure 4-8.





























Figure 4-8. Picture of PCB Model 480E09 Power Supply for ICP Laser Tachometer.

A suitable power supply for the angular rate sensor could not be found so it had to

be fabricated. The angular rate sensor requires a regulated power supply of 5 volts DC.

The decision was made to build the power supply using a voltage regulator integrated

circuit. Although the string potentiometer does not require a regulated power supply, it

was incorporated into the angular rate power supply so that both could be controlled

using a single DPST (double pole, single throw) switch. This will reduce the number of

switches that need to be switched when using the system. A wiring diagram of the power

supply, including the DPST switch, is provided in Figure 4-9.










ANGULAR RATE SENSOR STRING POTENTIOMETER


Figure 4-9. Wiring Diagram of Angular Rate Sensor Power Supply.

The power supply accepts +12 volts to provide power to the regulating circuit and + 24

volts to power the string potentiometer. A LM7805 voltage regulator chip is used to

provide a constant + 5 volts to the angular rate sensor. The 1000 tF capacitor helps to

maintain constant input to the regulator while the 0.1 gF capacitor eliminates any high

frequency pulses that may interfere with operation of the regulator. The LM7805 requires

an input voltage at least two to three volts greater than the output voltage. The car's

electrical system provides approximately + 12 volts to the regulator chip, which satisfies

the input voltage criterion. The + 24 volts is provided by a DC-DC converter normally

used to supply power for a laptop. The DC-DC converter accepts +12 volts and steps it

up to + 24 volts. The + 24 volts does not need to be regulated so a simple switch can be

used to turn it on and off. The LED is used to indicate the on/off status of the power

supply. Since the power supply receives power from the car's battery, it could drain the


ANGULAR RATE SENSOR


STRING POTENTIOMETER









battery if left on. A 3-pin connector is used to connect to each sensor to the power supply.

BNC connections are used for the outputs of the power supply for a quick a reliable

connection to the data acquisition system. All of the circuitry for the power supply is

contained in a plastic enclosure. The inputs to the power supply use 2-pin connectors so

the entire power supply can be removed or replaced quickly.

AC power is provided by a VectorMaxx 750-watt power inverter. The power

inverter is powered by the car's 12 volt electrical system. The output is a modified sine

wave with a frequency of 60 Hz and an amplitude of 115 volts (rms), similar to a

standard household outlet. The power inverter is connected to a Monster Cable brand

power strip for distribution of the AC power. All of the devices that use AC power are

connected to the power strip. The power inverter features short-circuit and overload

protection circuitry.

4.2.5 GPS

A Garmin GPS 16 was selected for the GPS receiver. The GPS 16 is WAAS

capable and uses the NMEA-0183 standard. It is enclosed in a rugged, waterproof

enclosure and features a magnetic bottom for quick installation. Since the laptop does not

have a serial port, a USB/Serial converter will be used to connect the receiver. The

converter also provides power to the receiver which cannot be done with a standard serial

port.

4.2.6 Video

The video system consists of four ultra-mini pinhole cameras (CCM630P), a quad

multiplexer (QC800), and a portable mini-DV recorder (Sony GVD-1000). The cameras

are color video cameras and measure 1.25 inches square. There is a choice of either a 780

or 900 viewing angle. The version with the 900 viewing angle was chosen because it









provided greater visibility for each camera. Each camera uses a 1/3-inch high-resolution

CCD sensor and features an automatic iris and backlight compensation. The cameras are

powered with 12 volts DC so the car's electrical system will be used to power them. A

fuse was installed for protection as well as a switch to turn the cameras off when not in

operation.

The quad multiplexer is a device most commonly used for security applications. It

combines the video signal from up to four cameras into one signal. There are several

options for displaying the four views: a large main view with the other three views shown

at the bottom, a large main view with a smaller window for a second view, a large main

view sequentially shifting between all cameras, and all four views shown at the same

time in equally sized windows. The latter option was selected since each view is equally

important.

A portable mini-DV recorder (Sony GV-D1000) was chosen for its exceptional

quality and ease of use. The recorder features a 4-inch LCD screen to verify the system is

working properly and cameras are aligned properly before recording. The mini-DV

format is a digital format with up to 500 lines of horizontal resolution. This is better

resolution than Hi-8 and S-VHS and it is digital. The digital format allows it to be

transferred to a computer easily for backup or editing. Some mini-DV tapes feature

memory chips that can be used to save information such as the date and time of the

recording. The memory chip can also be used to build a table of contents. The table of

contents contains a list of specific points in the recording. These points can be used for

easier access during playback or for editing to be done later.









The Quad multiplexer and mini-DV recorder use standard AC power and are

powered by the power inverter. Composite (RCA) video cables are used to connect the

cameras to the multiplexer and the multiplexer to the recorder. The mini-DV tapes can

record 90 minutes in EP mode. This is long enough to record the entire road course on a

single tape since the road course takes approximately 1 hour and 20 minutes to complete.

The entire video system is completely independent from the data acquisition system. The

only device the two systems share is the power inverter. The video will be used to further

review a subject's performance if necessary. It may also be used as a training tool for

new evaluators.

4.2.7 Software

National Instruments' proprietary software LabVIEW (version 6.1) will be used to

operate the data acquisition system. LabVIEW is a graphical programming language

designed specifically for interaction with National Instruments hardware. It features built-

in measurement and control subroutines, called virtual instruments, which allow for fast

and easy program development. A combination of several virtual instruments are wired

together to create a working program. A virtual instrument can be changed or created

from scratch to perform a custom operation. The user interface for the final program can

be configured to hide or show various controls (inputs) and indicators (outputs). The

program to be used for this system is fairly simple. It will be required to measure all of

the channels from the data acquisition system and record the data to the laptop hard drive.

National Instruments also provides the Measurement and Automatic Explorer (MAX)

which is used to configure the entire system. The National Instruments hardware

including the DAQ-Card, SCC, and SCC modules are entered in to MAX. Additionally,









MAX can be used to monitor each channel to ensure that each sensor is working

properly.

GPSLog software developed by U.S. Positioning LLC will be used to record the

GPS data. GPSLog is a simple program that records the output sentences from any

NMEA-0183 compliant GPS receiver. NMEA-0183 is a standard defining signal

requirements, data transmission protocols, and sentence formats for 4800-baud serial data

buses, which are widely used for GPS receiver communication (National Marine

Electronics Association). GPSLog can save either the raw data sentences or just

important information such as longitude, latitude, speed, and heading.

The MathWorks MATLAB software will be used to process and analyze the data.

MATLAB is a high-level technical computing language which is computationally faster

than traditional programming languages like C++ and Fortran (MathWorks Inc.).

MATLAB is the ideal application for inputting and analyzing the data. It features

toolboxes which contain subroutines for performing data analysis and statistics.

MATLAB is a very powerful application however it will be mainly used for simple

operations such as matrix manipulation and statistical analysis.

4.3 Installation

The system was installed in the vehicle in the least conspicuous way possible in

order to have a minimal impact on the driver. The driver is most likely driving a vehicle

different from their own and any equipment in plain view may make them nervous. The

majority of the equipment including the power inverter, SCC, power supplies, and video

equipment was installed in the trunk of the vehicle. A sheet of medium-density fiberboard

(MDF) was cut to fit in the trunk and was used as a mounting platform for the equipment.









A picture of the equipment installed in the trunk and a diagram showing the approximate

location of each component is shown in Figures 4-10 and 4-10 respectively.


Figure 4-10. Picture of Equipment Installed in Trunk.


Figure 4-11. Trunk Diagram Displaying the Approximate Location of Each Component.
(QUAD Video Multiplexer, PS Power Supply, ARS Angular Rate
Sensor, SP String Potentiometer, Acc. Accelerometer)


s ACC. PS

ScC ARS.SP PS ACC. PS


QUAD


POWER
LAPTOP INVERTER
MINI-DV TACH
PS

POWER
STRIP
CAMERA SWITCH









The laptop was originally installed in the trunk however the Occupational Therapists

requested it be located in the back seat so that it could be operated from inside the

vehicle.

There were few options for the locations of the laser tachometer and string potentiometer.

The laser tachometer had to be installed near either one of the front wheels because the

vehicle is front-wheel driven and does not have a rear axle. The string potentiometer

needed to be installed close to the steering shaft because it must be mechanically attached

to the shaft. A bracket included with the laser tachometer was fastened to an existing

mounting bracket for one of the car's brake lines. The laser tachometer is threaded and

was installed in the bracket using the provided hardware. The front axle was painted flat

black and a 1/4-inch wide piece of retroreflective tape was installed. A picture of the laser

tachometer setup is provided in Figure 4-12.

The tachometer was aimed at the retroreflective tape and the laser beam was

focused using the adjustment screw. When the sensor is powered, a LED illuminates to

indicate that the laser is being reflected and measured properly.

The string potentiometer is physically attached to the steering shaft via a pulley

and therefore had to be installed relatively close to the shaft. It was not possible to

remove the steering shaft in order to install an off-the-shelf pulley. The only feasible

option was to design a pulley that could be attached to the shaft without having to remove

it. The pulley is split in half so that it can be secured to the shaft using socket cap screws.

It features a groove to guide the cable from the string potentiometer as it winds around

the pulley. The steering shaft is not perfectly round and consequently the inside of the

pulley had a unique geometry. The pulley was machined out of a scrap piece of









aluminum using an end-mill and lathe. The best location that for the string potentiometer

that allowed the cable to reach the pulley unimpeded was on the lower part of the

firewall. The string potentiometer was secured to the firewall using J-B Weld two-part

epoxy. A picture of the string potentiometer and pulley assembly installed in the vehicle

is provided in Figure 4-13.


Figure 4-12. Picture of Laser Tachometer, Mounting Bracket, and Drive Shaft.


































Figure 4-13. Picture of Installed Sting Potentiometer and Custom Pulley Assembly.

A drawing of the custom pulley was generated using Pro Engineer modeling software and

can be found in Appendix C.

The accelerometers and angular rate sensor do not need to be located near any

particular area of the car. The main requirement is that they are attached securely to a

rigid part of the vehicle in order to measure accelerations and yaw accurately.

Furthermore, the three sensors should be located as close to the center of the car as is

feasibly possible. This is to reduce the amount of measured acceleration that may be a

result of suspension movement. The best location that would also allow for the sensors to

be out of sight was under either one of the front seats. The driver's seat features power

controls and as a result there are several motors underneath it. Therefore, the only option

was to install the sensors under the front passenger seat. A piece of carpet and insulating









foam was removed to allow access to the structural members of the vehicle. A frame rail

that ran laterally across the vehicle was used as a guide to position the longitudinal

accelerometer. The longitudinal accelerometer was located along the frame rail such that

it measured in a direction perpendicular to the frame rail. The orientation of the angular

rate sensor did not matter as long as it was positioned level so that it measured rotation

about the vertical axis of the car. It was positioned along the frame rail so that a

perpendicular face on the angular rate sensor could be used as a mounting location for the

lateral accelerometer. All of the sensors were attached using J-B Weld two-part epoxy. A

picture of the sensors installed beneath the passenger seat is provided in Figure 4-14.


























When the carpet is replaced and the passenger seat is in its normal position, the sensors
Iare compltFRAME RAIL


0 LONGITUDINAL
ACCELEROMETER





Figure 4-14. Picture of Lateral Accelerometer, Longitudinal Accelerometer, and Angular
Rate Sensor Installed Underneath the Front Passenger Seat.

When the carpet is replaced and the passenger seat is in its normal position, the sensors

are completely hidden.









The brake and turn signals were measured by wiring the positive and negative

(ground) leads of each in parallel with the inputs to their respective modules. All the

signals measure approximately 12 volts when in use. A logic statement was implemented

in LabVIEW that determines if the signal is above 6 volts. The result is a 1 corresponding

to the signal being activated and a 0 for it not being activated. The event trigger was

wired in the same was as the other signals except the positive 12 volts had to be supplied.

The event trigger is simply a pushbutton connected to the car's positive 12 volts and an

input to a module. The other input to the module is connected to the car's frame (ground).

The same logic used for the other signals was also used for the event trigger.

The GPS antenna was attached to the top of the trunk using the magnetic mount.

It was wired to a serial/USB converter which was connected to the USB port of the

laptop.

The positive lead of the power inverter was connected to the positive terminal of

the car's battery and the negative lead was connected to the vehicle chassis (ground). A

40-ampere fuse was installed on the positive lead near the battery to protect the inverter

and all of the equipment connected to it.

The multiplexer and mini-DV recorder was located in the trunk of the vehicle

with the rest of the equipment. The four cameras were installed in the ceiling of the

vehicle cab. They came with brackets which allowed them to be installed by screwing the

bracket into the rigid foam ceiling. The orientation of the cameras is completely

adjustable. Each one was wired with 12 volt DC power and a composite video cable. A

picture of the camera which monitors the driver's eye movements is provided in Figure 4-

15.






























Figure 4-15. Picture of Miniature Video Camera and Installation Bracket.

The video cable from each camera was wired to an input of the multiplexer. The output

of the multiplexer was wired to the input of the mini-DV recorder using a composite

video cable.

All of the wires for the various sensors, cameras, and power inverter were

concealed beneath the door sills, carpet, and headliner. The only evidence inside the

vehicle cab that would indicate the presence of special equipment is the four video

cameras. All of the cameras are mounted to the ceiling and although they are visible they

should not interfere with the driver's field of view. A wiring diagram of the entire system

is presented in Figure 4-16. The figure shows how the various components are connected

and the cables that were used. Figure 4-17 is also provided to show the location of the

components in the vehicle.








47




]--E- DACC

[ ] I ACC
LEGEND
[S] TACH -- POWER
G-- GROUND
ARS AC POWER
PS ------------ RCA VIDEO
STRING USB
POT o 68-PIN
COAX (BNC)
2V+4 3-RE
GND
I TRIGGER

O BRAKE

LEFT TURN

RIGHT TURN


Figure 4-16. Wiring Diagram of All System Components.








48















TERY

















LEGEND
-- POWER
-- GROUND
AC POWER
RCA VIDEO
USB
68-PIN
COAX (BNC)
3-W'RE


Figure 4-17. Wiring Diagram Indicating the Location of All System Components in the
Vehicle.









4.4 Calibration

The entire system except for the GPS receiver, event trigger, and brake and turn

signals was setup in a laboratory environment. The LabVIEW program used to operate

the system was designed and refined in the laboratory. The sensitivity for the

accelerometers provided by the technical specification sheets was verified using the

earth's gravity. Since they are static accelerometers, they will measure exactly 1 g when

placed on a level surface. There was no direct way of calibrating the angular rate sensor.

However, it was verified that the sensor measured 0 radians per second when stationary

and provided positive and negative voltage for clockwise and counterclockwise rotations

respectively. The laser tachometer and string potentiometer needed to be calibrated after

being installed in the vehicle. The laser tachometer was calibrated by driving at a

constant speed and comparing the result with the speed acquired by GPS. It was then

possible to calculate the factor which relates rotational frequency of the drive shaft in

hertz to vehicle speed in miles per hour. The string potentiometer was calibrated by

measuring the voltage with the wheel centered, turned one rotation to the left, and turned

one rotation to the right. There was no calibration needed for the event trigger, brake and

turn signals, or the GPS receiver.

4.5 Operation

A set of instructions was created such that the user could operate the system after

receiving minimal training. The first step is to turn on all of the equipment including the

power inverter, laptop, power supplies, and video equipment. Next, the user opens the

LabVIEW program and GPSLog. The data acquisition system is activated by pressing a

start button on the control panel of the LabVIEW program. GPSLog connects to the GPS

receiver and then starts logging the data. The video recording starts recording by pressing






50


record on the min-DV recorder. Once the system is running, the driving evaluation can

begin. At the end of the evaluation the programs are terminated and the video recording is

stopped. Finally, all of the equipment is turned off. The data from a one hour and twenty

minute evaluation is approximately 5 megabytes and is easily transferred to another

computer using a USB flash drive.














CHAPTER 5
SYSTEM VERIFICATTON

Once the system was installed in the vehicle, it was necessary to make sure that all

the components were working correctly and that the data provided was accurate. Before

implementing the system in actual research, a series of tests were performed to quantify

the capabilities of the system and build a model for analyzing the data. It was decided

that the data should be presented in a visual format so that the trends and interactions

between measurements could be easily observed. The next section discusses the

interpretation of the data and is followed by an analysis of the tests used to verify the

system's performance.

5.1 Interpretation of Data

As previously mentioned, the data needs to be represented in a visual format in

order to analyze the performance of the system. The independent variable for all the

measurements is time and it is plotted along the horizontal axis. The measurements are

made using a mixture of units so a conversion is necessary that will allow the full range

of each measurement to be graphed on a single plot. The values for acceleration

(measured in g' s) and yaw (measures in radians/second) were observed to be less than a

magnitude of one for most driving situations. Since there are three measurements whose

values are in the same range in their native units, the rest of the measurements are

normalized to fit in this range. The steering wheel position can be represented in radians,

degrees, or revolutions. The decision was made to measure steering wheel position in

revolutions because it is on the same order as the accelerations and yaw. Furthermore,









measuring it in revolutions allows an investigator to have a direct physical understanding

of the wheel position. Whether speed was measured in kilometers per hour or miles per

hour, it was too large to fit in the zero to one range. Since the FHWA research is focused

on the time before, during, and after an intersection, vehicle speeds will most likely be

less than 40 miles per hour. Therefore the units for speed are represented in miles per

hour divided by 20 (MPH/20). The brake and turn signals are simply on or off so they can

easily be represented in the zero to one range. In order to simplify the plots, the brake and

turn signal are only plotted when they are active.

In addition to being represented in different units, the different measurements are

also measuring quantities in different directions (e.g., longitudinal and lateral

acceleration). Therefore a sign convention must be developed so that the different

measurements can be plotted on a single graph and interpreted correctly. The sign

convention is based on the assumption that the vehicle is always moving in the forward

direction. Longitudinal acceleration will be positive for acceleration and negative for

deceleration. Speed will always be measured in a positive sense since it always had a

positive value. Lateral acceleration, yaw, and steering wheel position will be positive

when the vehicle is moving forward and making a right turn and negative when it is

moving forward and making a left turn. The brake status is positive when active and is

assigned a value of one simply to keep it out of the way of the other measurements on the

plot (i.e., non-zero). The turn signals will have a magnitude of 0.4 when active and will

be positive for right turns and negative for left turns. Basically, the measurements related

to forward movement (i.e., longitudinal acceleration and speed) will be the same sign

regardless of the direction of a turn. The turn-related measurements (i.e., lateral









acceleration, yaw, and steering wheel position) will be positive for right turns and

negative for left turns. The sign convention used when the vehicle is in reverse will not

be discussed here because there are not any reversing maneuvers in the FHWA research

study.

The GPS and vehicle dynamics data are recorded using two different programs that start

recording at different times. Therefore, the two sets of data must be synchronized. The

xcorr (cross-correlation) function in MATLAB is used to determine how much the GPS

data needs to be shifted in order to synchronize it with the vehicle dynamics data. The

laser tachometer has a lower threshold of about 4 miles per hour and cannot measure any

speeds lower than this value. Since the speed provided by the laser tachometer and GPS

are both sampled once per second, the GPS speed is used for all plots and calculations

since it provides the necessary data in the 0 to 4 miles per hour range. An example of a

maneuver plot for a left turn is provided in Figure 5-1. The abbreviations used for the

measurements and their units as depicted in the figure are listed in Table 5-1.

Table 5-1. Measurement Abbreviations and Respective Units for All Plots.
Measurement Abbreviation Units
Lateral Acceleration LA g
Longitudinal (Forward) FA g
Acceleration
Steering Wheel Position ST revolutions (off of
center)
Yaw YW radians/second
Speed (Laser Tachometer) SPT MPH/20
Speed (GPS) SPG MPH/20
Turn Signal Status TS ON/OFF
Brake Status BR ON/OFF







54


1.5
/ LA
---- FA
SST
1 YW
SPT
-- SPG
5' / SPG
+ TS
E 0.5 + TS
E BR
0 -






-0.5



1 I I I I I I I I
500 505 510 515 520 525 530 535 540 545
time [sec]

Figure 5-1. Example Plot of a Typical Left Turn.

Figure 5-1 shows all of the measurements represented on a single plot. In order to better

understand and interpret the values of the results, several simple driving maneuvers were

performed. All of the maneuvers were performed in an empty parking lot with lane

markings. The four maneuvers were (1) accelerating from a stop, (2) decelerating to a

stop, (3) 90-degree right turn, and a (4) 90-degree left turn. Each maneuver was

performed in two distinct manners. The maneuver was performed once in a normal, safe

manner. The maneuver was then performed a second time in an aggressive manner,

almost pushing the vehicle to its limits.

5.2 Analysis of Acceleration Test Maneuver

For the acceleration test maneuvers, the vehicle was at rest and accelerated until the

vehicle reached 25 miles per hour. For the normal driving condition case, the accelerator

pedal was gently depressed and the car accelerated gradually. For the aggressive driving









condition, the accelerator pedal was abruptly applied and the car was accelerated quickly

but not enough for the tires to lose traction. The plots for the normal and aggressive

acceleration test maneuvers are provided in Figures 5-2 and 5-3 respectively. For the rest

of the analyses in this chapter, all of the plots will use the speed measured with GPS and

will not display turn signal data.

2 I----------------------
LA
---- FA
1.5 ST
YW
SSP
Q x BR



S0.5
--"- -'= ------------,-, ---- -







-0.5



452 454 456 458 460 462 464
time [sec]

Figure 5-2. Plot of a Typical Normal Acceleration Test Maneuver.

From the plot of the normal acceleration, it can be seen the brake is initially applied

and then released when the car begins to accelerate. The lateral acceleration, yaw, and

steering wheel position are negligible as expected during a straight-line maneuver. The

longitudinal acceleration reaches a maximum value of 0.18 g and then reduces gradually.

When comparing the two scenarios, the aggressive case achieves the top speed much

quicker and has a higher longitudinal acceleration of 0.40 g. The time needed to reach 25







56


MPH was roughly 8 seconds for the normal case and 4.5 seconds for the aggressive case.

The results from these tests are as expected and help to quantify the car's performance. If

a driver were to accelerate with an acceleration of 0.4 g then it can be assumed that the

driver is driving aggressively and at the car's limits.





---- FA
ST
E- YW
S 1 SP
E x BR

-5 0.5






-0.5 -
-0







537 538 539 540 541 542 543 544 545 546
time [sec]

Figure 5-3. Plot of a Typical Aggressive Acceleration Test Maneuver.

5.3 Analysis of Deceleration Test Maneuver

A similar analysis was performed for the deceleration test. The car was decelerated

in a straight-line from about 25 MPH. The plots for the deceleration test maneuvers are

provided in Figures 5-4 and 5-5. The normal deceleration generated a maximum force of

about -0.19 g while the aggressive deceleration reached nearly -0.43 g. The normal case

took approximately 6 seconds to decelerate to a stop from 20 MPH. On the other hand,

the aggressive case required only 3.5 seconds. Again, these results are as expected and

assist in quantifying the deceleration capabilities of the car.
















1.5 -














-----_---_ .. --
S1








1



-0.5




496 498 500 502 504 506 508
time [sec]

Figure 5-4. Plot of a Typical Normal Deceleration Test Maneuver.






1.5



C1




-0.5


0...

,. .


-1 I I I I I I I1
573 574 575 576 577 578 579 580
time [sec]

Figure 5-5. Plot of a Typical Aggressive Deceleration Test Maneuver.










5.4 Analysis of Right Turn Test Maneuver

The right and left turn test maneuvers were performed in a similar manner. The

same roadway lane markings were used to ensure the geometry of the turns is identical.

In all cases, the driver approached the turn, applied the brake, and then completed the

turn. The only difference between the normal and aggressive scenarios was the speed at

which the turn was taken. The speed during the midpoint of the turn was about 10 MPH

for the normal case and 15 MPH for the aggressive case. The turns are tight and resemble

those of a residential neighborhood. So, taking one of the turns at 15 MPH was actually

close to the car's limits. The plots for the normal and aggressive right turn test maneuvers

are provided in Figures 5-6 and 5-7 respectively.



LA
---- FA
1.5 ST
YW
SP


E

S 0 .5 .



0 .
Or ; r--''----------' '


-0.5



774 776 778 780 782 784 786 788
time [sec]

Figure 5-6. Plot of a Typical Normal Right Turn Test Maneuver.







59




-LA
---- FA
1.5 ST
YW
SP
1 x BR














820 821 822 823 824 825 826 827 828 829 830
time [sec]

Figure 5-7. Plot of a Typical Aggressive Right Turn Test Maneuver.

In both cases, the longitudinal acceleration and speed relationship is as expected

based on the acceleration and deceleration test cases. The speed decreases when the

longitudinal acceleration is negative and the brake is applied. The speed begins to rise at

the end of the turn when the longitudinal acceleration is positive. The most relevant

measurements for the turning maneuvers are lateral acceleration, yaw, and steering wheel

position. The profiles for the three measurements are similar however the magnitudes of

lateral acceleration and yaw differ significantly. The steering wheel position reaches a

maximum value of 0.695 revolutions for the normal case and 0.825 revolution for the

aggressive case. These two values should be approximately the same because the

geometry of the turn is the same for both cases and consequently the path traveled by the

vehicle should also be the same. The reason for the difference in the steering wheel

position values can be determined by examining the profiles. For the normal case, the









profile looks nearly symmetric and is within 10 percent of the maximum value for 1.6

seconds. For the aggressive case, the steering wheel position reaches its maximum value

more abruptly and is within 10 percent of this value for only 0.9 seconds. Therefore, it

can be inferred that for the normal case the driver turned the wheel gradually as they

entered the turn, then held the wheel in a fixed position during the turn, and finally

gradually turned the wheel back to center at the exit of the turn. On the other hand, for the

aggressive case, the driver turned the wheel abruptly to begin the turn and almost

immediately began returning it to center. Since the aggressive turn was completed at a

higher speed, it is expected that there will be higher values for yaw and lateral

acceleration. This can be verified by examining the plots. The lateral acceleration reaches

a maximum value of 0.332 g for the normal case and 0.561 g for the aggressive case.

Yaw reaches a maximum value of 0.531 radians/second for the normal case and 0.685

radians/second for the aggressive case. A similar discussion will now be presented for the

left turn test maneuvers.

5.5 Analysis of Left Turn Test Maneuver

The results from the left turn test maneuvers are nearly identical to those from the

right turn test maneuvers. The plots for the normal and aggressive left turn test

maneuvers are provided in Figures 5-8 and 5-9 respectively. The same trends that were

observed for the right turns test maneuvers are also observed for the left turns. The

steering wheel position reaches similar maximum values of 0.71 and 0.77 revolutions for

the normal and aggressive cases respectively. The maximum lateral acceleration achieved

0.248 g for the normal case and 0.603 g for the aggressive case. Yaw reached a maximum

value of 0.423 radians/second for the normal case and 0.682 radians/second for the

aggressive case.





















0.5


-0.5 E


-1
666


672
time [sec]


Figure 5-8. Plot of a Typical Normal Left Turn Test Maneuver.


0.5


-0.5 I


-71
71


3


716 71
time [sec]


Figure 5-9. Plot of a Typical Aggressive Left Turn Test Maneuver.


- LA
---- FA
ST
YW
- SP
x BR


t t~


". --.LA.


__ LA
---- FA
ST
YW
SP
SBR


I I I I I I


*









These results are as expected and confirm the results from the right turn test

maneuvers. It can be assumed that if a driver achieves a lateral acceleration value near

0.6 g for any driving maneuver, they are driving the car at its limits. However, a similar

statement cannot be made for the yaw value. This is because yaw depends on the

curvature of the turn (i.e., geometry of the intersection). Furthermore, the results for right

and left turns measured on a road course will be different because in general right turns

will have a smaller radius than left turns.

This analysis provides a basis for understanding the results obtained from the road

course evaluations. Most importantly, the capabilities of the test vehicle have been

quantified in terms of longitudinal and lateral acceleration. If a driver reaches these

limits, then they are most likely driving in an unsafe manner. Although this was a general

analysis, a more in-depth analysis could be performed for any particular driving

maneuver depending on the hypothesis being tested. For example, if a research study was

focused on driving behaviors during lane changes, then several lane changes could be

performed at different rates and vehicles speeds for a better understanding of the

maneuver.














CHAPTER 6
FHWA RESEARCH ANALYSIS

This chapter discusses the procedures used to analyze the data for the FHWA

research study. The purpose of the FHWA research is to study the effectiveness of

intersection improvements recommended by the FHWA. Participants navigate a road

course and their driving performance is evaluated for ten intersections. The ten

intersections are five pairs of similar intersections where one intersection features the

recommended improvement and the other one does not. The naming convention for the

intersection uses the format 1A, JB, 2A, 2B, etc. where the number refers to the

intersection pair and A and B represent improved and unimproved intersections

respectively.

The analysis of the road course data is conducted in three separate steps and there is

a MATLAB M-file specific to each step. The MATLAB files are named main, stats, and

sort respectively. The main file is used to locate the ten intersections using the GPS data.

It also converts the voltage measurements from each sensor to their desired units and then

creates an output file for each maneuver (intersection). The stats file is used to separate

each maneuver into three phases (approach, turn, and recovery). It calculates statistics for

each measurement of each phase of each maneuver. The statistics are compiled and

output files are generated for each of the three phases. Lastly, the sort file takes all of the

statistic files from each subject and sorts the data by maneuver rather than subject. The

final result is fifteen files which contain the statistics for each phase of each intersection

pair. Finally, SPSS for Windows statistics software is used to analyze the results and









compare the improved and unimproved intersections. All of the MATLAB codes for this

chapter may be found in Appendix D.

An almost identical analysis is performed on the simulator data. There are ten

simulator scenarios that are replicas of the ten road course maneuvers. The simulator

creates output files that are similar to the output files generated by the main program used

to analyze the road course data. Similar versions of the stats and sort programs are used

to analyze and format the data so that is in a form identical to that of the road course data.

The exact same procedure is used to analyze the statistics from the road course and

simulator using SPSS. A more detailed discussion of each MATLAB file and the SPSS

analysis will now be presented.

6.1 Description of main.m (Road Course)

Before the file can be run, the data must first be entered into the MATLAB

workspace. The vehicle dynamics data and the GPS data are both stored in comma-

delimited text files. The GPS data contains a header and text stings (timestamp from

GPSLog) so MATLAB's Import Data Wizard is used to add the data to the workspace.

The Import Data Wizard recognizes text and automatically sorts out the numerical data to

be imported. The vehicle dynamics data is imported using the dlmread command. The

first step in the program is to plot the route traveled by the vehicle using the GPS

coordinates. A simple algorithm is implemented to determine when the vehicle is at the

start or end location for a particular intersection. The desired start and end locations were

chosen by the driving evaluators. Their GPS coordinates were found by driving the road

course and marking each location by using the event trigger. A plot of the entire road

course from GPS coordinates is shown in Figure 6-1.












29.71 -


29.7 -


D29.69 3A C 11A




29.67 -
B


29.66 76A 4A 4B


29.65
-82.41 -82.4 -82.39 -82.38 -82.37 -82.36 -82.35 -82.34
Longitude [Decimal Degrees]


Figure 6-1. FHWA Road Course Generated Using GPS Coordinates with Intersections of
Interest Highlighted and Labeled.

The ten intersections for which drivers are evaluated are highlighted and labeled at the

start of each maneuver. This plot is generated to ensure that the proper intersections are

selected for further analysis.

The vehicle dynamics data and GPS data are recorded using two different

programs and consequently the two sets of data are not synchronized. A correlation

analysis between the speed provided by GPS and the speed provided by the laser

tachometer is performed. The correlation is used to find the time offset between the two

sets of data and the GPS data is adjusted accordingly. A plot of each maneuver is

generated to verify that the GPS data was adjusted correctly. An example of one of these

plots is provided in Figure 6-2.







66


Maneuver 6B


505 510 515 520 525
time [sec]


530 535 540 545


Figure 6-2. Example of Maneuver Plot Used to Verify That The Vehicle Dynamics Data
and GPS Data Are Synchronized.

If for some reason the two sets of data are not synchronized exactly, the user has the

option to shift the GPS data by the desired number of seconds. The plots are regenerated

and the process is repeated until the data is synchronized.

The final step is to create output files for each maneuver. The output files contain

all of the measurements in their appropriate units for the entire duration of each

maneuver. The files are created using the dlmwrite command in MATLAB. The result is

a comma-delimited file for each maneuver which can easily be imported into MATLAB

for further analysis.


1.5



1

C0

E 0.5



.. 0
,...









6.2 Description of stats.m (Road Course)

The stats MATLAB file is responsible for determining the three phases

(approach, turn, and recovery) of each maneuver and calculating statistics for each phase.

The first step is to import the data by reading the files generated by the main file using

the dlmread command. The next step is to determine the three phases of each maneuver.

This is done by implementing an algorithm which uses the heading from GPS and vehicle

yaw. The algorithm uses the heading information to determine when the vehicle is in the

middle of the turn. It then analyzes the yaw data to find out when the yaw is less than

0.05 radians per second. Since the yaw is near its maximum during the turn, the last point

before the turn when the yaw is still less than 0.05 radians per second is considered to be

the start of the turn. The first point after the turn when the yaw returns to less than 0.05

radians per second is considered to be the end of the turn. The point immediately before

the start of the turn is considered to be the end of the approach phase and the point

immediately after the end of the turn is considered to be the start of the recovery phase.

The start of the approach phase and the end of the recovery phase are determined by the

start and end locations of the maneuver respectively. Although the distances for the

approach and recovery phases vary from maneuver to maneuver, they will always be the

same for the same maneuver because the GPS locations for the start and end of the

maneuver are constant. For comparison with the simulator scenarios, the distances of the

approach and recovery phases need to be equal to those of the simulator. This is

accomplished by converting the latitude and longitude to feet and then calculating the

distance traveled during each phase. The approach and recovery phases can then be

truncated accordingly to equal the distances traveled in the respective simulator

scenarios.







68


The stats program generates a plot for each maneuver so the user can verify that

the turn phase was selected correctly. A large and bold cross is used to mark the start and

end of the turning phase. An example of one of these plots is shown in Figure 6-3.


Maneuver 4B (Subject XX)
1.5- '-'




-0.5 --FA
5 '- -.'










Tire [sec]r
-LA

-0.5 : .. FA
.. ST

.. SP

1565 1570 1575 1580 1585 1590 1595 1600 1605 1610
Time [sec]


Figure 6-3. Plot of Maneuver 4B Generated Using stats.m With Markers Indicating the
Start and End of The Turning Phase.

Once the selection of phases is verified, the program is used to calculate statistics for

each phase of each maneuver. The twelve statistics calculated for each measurement are

minimum, maximum, mean, root-mean-square, integral, variance, median, range,

skewness, kurtosis, sum/time, and sum/distance. The last two statistics listed are the sum

of the respective measurement normalized by the time and distance traveled respectively

during the phase. These twelve statistics are calculated for lateral acceleration,

longitudinal acceleration, steering, yaw, speed, and combined acceleration. Combined









acceleration is the absolute magnitude of the overall acceleration of the vehicle and is

defined in Equation 6.1

co accel = cce accel2 + long accel Eq. 6.1

The final result is 72 statistics for each phase. Since there are three phases per maneuver,

and a total often maneuvers, there are 2,160 statistics calculated for each subject. These

statistics are compiled into comma-delimited text files using the dlmwrite command in

MATLAB. There are three files, one for each phase, that contain the statistics for all ten

maneuvers. In addition to these statistics, the user enters the subject's age and gender.

The age is used to determine what age group the subject is in depending if the driver is

young or old (65 years and older). The age group, gender, and maneuver number is also

included in the output files to assist with the SPSS analysis.

6.3 Description of sort. m (Road Course)

The purpose of the sorting program is to compile all of the statistic files and sort

them by maneuver pairs. The program reads all of the statistic files from each subject and

arranges them by maneuver number. The program generates fifteen comma-delimited

text files. There is one file for each phase of each maneuver pair. In each file there is a

row for each subject. The row contains subject number, age group, gender, and maneuver

number. It is then followed by all of the statistics for the improved maneuver (i.e., A)

which is followed by all of the statistics for the unimproved maneuver (i.e., B). The

statistics are arranged in this order so that they can easily be analyzed using SPSS.

6.4 Description of sinmm and sim_sort.m

The simulator software produces files nearly identical to those produced by the

main.m program for the road course. A program titled sim.m is very similar to stats.m for









the road course and calculates statistics for each phase of each maneuver. There are

subtle differences between the two programs. For example, the simulator data is provided

in different units compared to those used for the road course, so sim.m is used to convert

the data to the appropriate units. Also, the algorithm used to determine the phases of each

maneuver is slightly different due to the fact that the heading for the simulator always

goes from zero to plus (right turn) or minus (left turn) ninety degrees. This is different

than the road course heading which is the actual global heading of the vehicle and varies

based on the orientation of the intersection. The program generates statistics files which

are identical to those generated by the stats.m program for the road course.

A sorting program named sim sort.m is used to sort the simulator statistics files in

much the same way as sort.m did for the road course statistics. There are two minor

changes that were made. First, the order of the maneuvers in the statistic files for the

simulator is different from the order for the road course, so the indexing used to sort them

is changed appropriately. The other change is to account for an extra intersection named

IC. This intersection is similar to lB except that is has a different improvement feature

recommended by the FHWA. This improvement could not be found in any intersections

in Gainesville, FL and as a result could not be incorporated into the road course. The

effect of the improvement wanted to be studied so a simulator scenario was created. The

sim sort.m program creates fifteen output files similar to those created for the road

course. The only difference is that the files for the maneuver 1 intersections contain the

data for three maneuvers instead of two.

6.5 Description of SPSS Analysis

SPSS statistical analysis software is used to analyze the results from the road

course vehicle and simulator statistics. A general lineal model (GLM) repeated measures









analysis was performed. This provides analysis of variance when the same measurement

is used several times on each subject or case. For this analysis, the 1I i/thin-%ulbje, t variable

is the improvement of the intersection. The between-subject variable is age and it is used

to divide the population into two groups (young and old). This general linear model

analysis can test the effects of both the between-subject factors and the ii i/thin-%Iilje. t

factors as well as the interaction between the two factors. The specific options chosen for

the analysis were descriptive statistics, estimates of effect size, and observed power.

Additionally, a plot of the estimated marginal means of a particular measure is generated.

This plot has a line for each age group, a vertical axis for the mean of the measurement,

and a horizontal axis for the improvement status of the intersection. The plot provides a

useful too for visually analyzing the effects of each factor and the interaction between

them. A sample output file for one phase of one maneuver and five statistics (maximum

lateral acceleration, maximum longitudinal acceleration, maximum combine acceleration,

maximum yaw, and maximum speed) is provided in Appendix E. Clearly performing and

analyzing the results of this analysis for all statistics of all phases of all maneuvers would

be extremely time consuming. The analysis calculates the significance of the effect of

each factor on a statistic and is used to determine which statistics are different as a result

of intersection improvement or age group. Once some of the significant statistics are

found, they can be further analyzed to infer about the effect of the intersection

improvement or age group. For example, if both age groups had a higher maximum speed

for the improved intersection, it could be inferred that the improvement helped by

allowing the majority of participants to achieve a higher speed during the turn. This entire






72


procedure reduces the time needed for analysis by analyzing a particular statistic for all

subjects at the same time rather than on an individual basis.














CHAPTER 7
RESULTS AND DISCUSSION

7.1 FHWA Research Results

At the time of this thesis submission, the FHWA research study had recently

concluded. It is currently in the initial stages of data analysis. The turning phase was the

first phase to be analyzed because it captures the dynamics during the intersection which

is the main focus of interest. The approach and recovery phases will be analyzed at a later

date to determine driver stability before and after the intersections. There are 72 statistics

calculated for each phase, but it would be too time consuming to analyze all of the

statistics for each maneuver using SPSS. The first statistics to be investigated for the

turning phase of all maneuvers were (1) maximum combined acceleration, (2) maximum

longitudinal acceleration, (3) maximum lateral acceleration, (4) maximum yaw, and (5)

maximum speed. SPSS can be used to determine if these statistics are significantly

different for older and younger drivers (age group), improved and unimproved

intersections (intersection improvement), and the interaction between age group and

intersection improvement. An example analysis of these five statistics for one of the

FHWA road course maneuvers will now be presented.

7.2 Example Analysis of Turning Phase of FHWA Maneuver

The five statistics are compiled into tables for each maneuver. The average and

standard deviation for each statistic is tabulated for four categories. The four categories

are improved-young, improved-old, unimproved-young, and unimproved-old, where

improved/unimproved refers to the intersection and young/old refers to the driver's age










group. The statistics for Maneuver 2 from the FHWA road course are presented in Table

7-1.

Table 7-1. Summary of Statistics for Turning Phase of FHWA Road Course Maneuver 2.
Maximum Maximum Maximum Maximum Maximum
Combined Longitudinal Lateral Yaw Speed (mph)
Acceleration Acceleration Acceleration (radians/sec)
(g) (g) (g)

Descriptive statistics
Intersection
Age Ave SD Ave SD Ave SD Ave SD Ave SD

Improved
Improved 0.30 0.05 0.19 0.06 0.25 0.06 0.40 0.06 22.17 2.66

Old 0.30 0.06 0.19 0.08 0.22 0.06 0.37 0.07 20.75 2.91
Unimproved Young 0.33 0.05 0.19 0.04 0.31 0.05 0.46 0.05 19.48 4.47
Old 0.33 0.05 0.18 0.04 0.32 0.06 0.46 0.05 18.44 6.30

Inferential Statistics

F p F p F p F p F p
Age Group
(Older Versus 0.12 0.73 0.11 0.74 0.49 0.49 1.29 0.26 3.03 0.09
Younger Drivers)
Intersection Type
(Improved versus 19.5 0.01* 0.21 0.65 71.50 0.01* 59.33 0.01* 11.53 0.01*
Unimproved)
Interaction
0.51 0.47 0.17 0.68 4.45 0.04* 2.04 0.16 0.07 0.80
(Age x Intersection)

Maneuver 2 of the FHWA road course is a 90-degree right turn onto a multi-lane divided

highway. Both the improved and unimproved intersections are signalized. The improved

intersection features right-turn channelization. The right-turn channelization is a

dedicated lane for all vehicles that are making a right turn at the intersection. It extends

all the way into the receiving highway and allows drivers to accelerate up to speed before

merging into traffic. A picture of the right-turn channelization is provided in Figure 7-1.































"-". ..
S a m .. ..... "-







Figure 7-1. Picture of Right-Turn Channelization from FHWA Road Course Maneuver
2A.

An analysis of the statistics for FHWA Road Course Maneuver 2 will now be

presented. First, the significance, if any, of each variable is investigated. There are four

measurement statistics that are significant (p < 0.05) for the type of intersection

(improved/unimproved). They are maximum combined acceleration, maximum lateral

acceleration, maximum yaw, and maximum speed. The improved intersection has lesser

values for all of these statistics except maximum speed. This can be explained by

examining the equations for lateral acceleration and yaw which are provided in Equations

7-1 and 7-2 respectively.

2
a, = Eq. 7-1
r










co =- Eq. 7-2
r

where v is the vehicle velocity and r is the radius of curvature.

It should be noted that these equations are derived from their governing differential

equations and in their current form are valid only for an arc of constant curvature.

However, they can still be used to understand the interaction between speed, lateral

acceleration, and yaw. Since the speed is greater for the improved intersection and the

yaw and lateral acceleration is less, it can be inferred that the radius of curvature is

greater than for the unimproved intersection. This can be easily verified by examining the

geometry of each intersection. The results indicate that drivers are able to negotiate the

improved intersection at higher speeds while experiencing less lateral forces.

Maximum lateral acceleration is the only variable that is significant for the

interaction between age group and intersection type. The older drivers had a greater value

of maximum lateral acceleration for the unimproved intersection and a smaller value for

the improved intersection. Since the maximum speed is greater for the younger age group

for both intersections, we would expect the respective values for maximum yaw to also

be greater. The only explanation is that the maximum statistic measures do not indicate

the time at which they occurred. If it is assumed that both age groups of drivers achieved

maximum lateral acceleration at approximately the same time (midway through the turn).

Then it is likely that the younger and older drivers reached their maximum speed at

different times. Further analysis of the data could reveal if the older drivers began to

decelerate earlier.

Maximum speed is the only variable that approaches significance (p < 0.10) for

the age group. Younger drivers had greater speed for both the improved and unimproved









intersections. This is an expected result since older drivers generally drive with more

caution and at a lower speed.

This analysis, or one very similar can be performed on the remaining maneuvers.

Examination of all 72 measurement statistics can reveal which ones are most significant.

The measurement statistics of interest may vary depending on the specific improvement

being examined in each maneuver. Furthermore, the statistics can be separated into

different cases depending on the driving conditions (i.e., stopping before the turn or

signal status). The other phases (approach and recovery) of each maneuver will also be

analyzed using a similar process. The approach and recovery phases will be analyzed to

determine how each intersection affects the driver's stability before and after the turn. It

has been shown that the measurement statistics and their associated significance is useful

in analyzing each intersection. These results combined with an understanding of the

intersection can be used to infer information about the driver's performances.














CHAPTER 8
CONCLUSION

The vehical dynamics acquisition system developed for the FHWA study was able

to successfully measure and record the vehicle dynamics of a Buick sedan. The addition

of GPS allowed the system to acquire the vehicle dynamics at specific locations. The

system is flexible and can measure a wide variety of inputs. It can easily be expanded

with the addition of SCC modules and transducers to measured additional information.

The system was first used for the FHWA study and proved to be a useful tool in

quantifying driving behavior. It could be used in its current form for a different study

concerned with a different driving behavior. The only change would be to the MATLAB

code in order to analyze the particular area of interest. The video equipment, although not

directly incorporated with the data acquisition and GPS equipment, was a helpful tool for

the evaluators. It was used to train new evaluators and to review evaluations to ensure

that evaluators were scoring driving errors consistently.

8.1 Future Research Studies

It is likely that this system will be used in the future for other driving behavior

studies. If needed, additional sensors can easily be added. For example, a load cell could

be used to measure the amount of force the driver is applying to a pedal. The system

could even be adapted to measure physiological responses from the driver such as heart

rate, or perspiration. Also, the cameras can be remounted to change the field of view. For

example, the cameras could be used to monitor vehicle lane maintenance by aiming the

cameras toward the lane markings next to the vehicle.









8.2 Recommendations and Improvements

Although the system worked well and as intended, some improvements could be

made that would make it easier to use and operate. The first recommendation is to find an

alternate way to measure the vehicle speed. The speed obtained from the GPS was only

updated once per second and was less accurate when the vehicle was stationary or

moving slowly. The speed measured using the laser tachometer was inconsistent due to

the laser not being reflected sometimes. Furthermore, the laser tachometer could not

measure speed below 5 miles per hour. An infrared ground speed sensor like the

Correvit LFTM (Appendix B) would be able to measure vehicle speed faster and more

accurately.

The second recommendation is to use only one program for operating the system.

This could be done by incorporating the GPS with the rest of the system via a serial port

and LabVIEW. The LabVIEW Virtual Instrument could be programmed to interface with

the GPS antenna using the NMEA 0183 Standard.

The final recommendation is related to the overall design and installation of the

system. If a similar system were to be installed in another vehicle then all of the

equipment should be contained in a single container or rack. The rack would contain the

power inverter, all power supplies, computer, and data acquisition system. There should

be one panel that contains all of the connections for the sensors. Also, a personal

computer (PC) could be used instead of a laptop computer. This would allow the

hardware to be contained in the rack and the monitor located in the cab of the vehicle.

The computer could be controlled using a Bluetooth keyboard and/or mouse. The

advantage of designing the system like this is that it would take up less room in the

vehicle, be simpler to operate, and easier to install or remove.















APPENDIX A
INTERSECTION DIAGRAMS






















































22 DEC. 80 N.W. 13TH STREET & N.W. 34TH STREET 342






81


































NW 53RD


MOOE 371







83








1 01 N.W. 39TH AVENUE



02---



- r N 005 -0
01 J 4& ^


-06 -







-i 404
I 7-- NODE
DRVWIHG UPDATE: /3/96 07 L_

V1& 0 6 04&07




-OR )02&06 0 &
04&08
"------ 12"
R ORRO






I I i

216 02&05 03&HV8

26 FEB. 90 N.W. 34TH STREET & N.W. 39TH AVENUE 340


















































































II APRIL 1978 N. STH AVENUE & N.W. I6TH TERRACE 316


~ ~II~
























N.W. 8th AVENUE






02_ |


1 ~~~r w -Ii-


I-










C,
To
C

34
z1


O'2


NODE 0443


19 MAY 1978 N.W. 8TH AVENUE a N.W. 22ND STREET 317













































NODE 0437


EXCLUSIVE
PED BY
T.O FD.


u-I
z


I~






-ii







88







i *












002







S NEWBERRY RD.
-LOOP L N SIDEWALK
LOOP IN SIEWAL NODE 00675
; ACCESS RD. 13265
< o

/t i w

01 6 OL-A fewrI4 2 LrZ & -A' 9L4












0- OR 1 -





19 DECEMBER '89 NEWBERRY RD. (SR26) a N.W. 55TH STREET 133




Full Text

PAGE 1

VEHICLE DYNAMICS MEASUREMENT SYSTEM FOR EVALUATING OLDER DRIVER PERFORMANCE By ETHAN J. DAVIS A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ENGINEERING UNIVERSITY OF FLORIDA 2006

PAGE 2

Copyright 2006 by Ethan J. Davis

PAGE 3

This document is dedicated to my gr andfather Robert ‘Sport’ Davis Sr.

PAGE 4

iv ACKNOWLEDGMENTS I would like to thank my advisor Dr. Niezreck i for assisting me with this research and providing useful knowledge and guidance. I would also like to thank Dr. John Schueller for helping me with many on-campus i ssues and Dr. Carl Crane for serving on my supervisory committee. Furthermore, I woul d like to express gratitude to everyone at the Occupational Therapy Department at th e College of Public Health and Health Professions. I would also like to thank Richard Phillips for he lping me with parts of my research and providing some helpful tips along the way. Finally, I would like to thank my parents and the rest of my family for their continued support.

PAGE 5

v TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES...........................................................................................................viii LIST OF FIGURES...........................................................................................................ix ABSTRACT....................................................................................................................... xi CHAPTER 1 INTRODUCTION........................................................................................................1 1.1 Background.............................................................................................................1 1.2 Older Driver Statistics............................................................................................2 1.3 Purpose of FHWA Research...................................................................................3 1.4 Objectives...............................................................................................................5 2 LITERATURE REVIEW.............................................................................................9 2.1 Previous Research Studies......................................................................................9 2.2 Devices.................................................................................................................13 2.2.1 DriveCam...................................................................................................13 2.2.2 Vericom......................................................................................................14 2.2.3 Dewetron....................................................................................................16 3 SYSTEM REQUIREMENTS.....................................................................................18 3.1 Performance..........................................................................................................18 3.2 Transducers...........................................................................................................19 3.3 GPS.......................................................................................................................2 0 3.4 Video.....................................................................................................................2 1 3.5 Ease of Use...........................................................................................................21 3.6 Cost....................................................................................................................... 21 3.7 Flexibility..............................................................................................................22 4 SYSTEM DESIGN.....................................................................................................23 4.1 Vehicle Description..............................................................................................23

PAGE 6

vi 4.2 Selection of Components......................................................................................26 4.2.1 Data Acquisition System............................................................................26 4.2.2 Transducers.................................................................................................28 4.2.3 Modules......................................................................................................31 4.2.4 Power Supplies...........................................................................................32 4.2.5 GPS.............................................................................................................36 4.2.6 Video..........................................................................................................36 4.2.7 Software......................................................................................................38 4.3 Installation............................................................................................................39 4.4 Calibration............................................................................................................49 4.5 Operation..............................................................................................................49 5 SYSTEM VERIFICATTON......................................................................................51 5.1 Interpretation of Data............................................................................................51 5.2 Analysis of Acceleration Test Maneuver.............................................................54 5.3 Analysis of Deceleration Test Maneuver.............................................................56 5.4 Analysis of Right Turn Test Maneuver................................................................58 5.5 Analysis of Left Turn Test Maneuver..................................................................60 6 FHWA RESEARCH ANALYSIS..............................................................................63 6.1 Description of main.m (Road Course)..................................................................64 6.2 Description of stats.m (Road Course)...................................................................67 6.3 Description of sort.m (Road Course)....................................................................69 6.4 Description of sim.m and sim_sort.m ...................................................................69 6.5 Description of SPSS Analysis..............................................................................70 7 RESULTS AND DISCUSSION.................................................................................73 7.1 FHWA Research Results......................................................................................73 7.2 Example Analysis of Turning Phase of FHWA Maneuver..................................73 8 CONCLUSION...........................................................................................................78 8.1 Future Research Studies.......................................................................................78 8.2 Recommendations and Improvements..................................................................79 APPENDIX A INTERSECTION DIAGRAMS.................................................................................80 B COMPONENT TECHNICAL SPECIFICATIONS...................................................92 C CUSTOM PULLEY DRAWINGS.............................................................................95 D MATLAB CODES.....................................................................................................97

PAGE 7

vii E EXAMPLE OF SPSS OUTPUT FI LE FOR MAXIMUM YAW DURING THE TURN PHASE OF MANEUVER 1.........................................................................118 LIST OF REFERENCES.................................................................................................120 BIOGRAPHICAL SKETCH...........................................................................................122

PAGE 8

viii LIST OF TABLES Table page 4-1 Technical Specifications for Signa l Conditioning Analog Input Modules..............32 4-2 Transducer Output Range s and Module Input Ranges.............................................32 5-1 Measurement Abbreviations and Respective Units for All Plots.............................53 7-1 Summary of Statistics for Turning Ph ase of FHWA Road Course Maneuver 2......74

PAGE 9

ix LIST OF FIGURES Figure page 1-1 Street Map of The FHWA Road Course in Gainseville, FL......................................4 1-2 Picture of NODRTC Simulator showi ng vehicle cab and projection screens............7 2-1 Display Output from ARGOS Instru mented Vehicle (Rizzo et al., 292)................11 2-2 Picture of DriveCam.................................................................................................14 2-3 Picture of Vericom VC3000.....................................................................................15 2-4 Picture of Dewetron DEWE5000.............................................................................16 4-1 Exterior View of 2005 Buick Centur y Sedan NODRTC Research Vehicle............23 4-2 Picture of Left Foot Accelerator Pedal Adapter.......................................................24 4-3 Picture of Hand Controls, Turn Signal Extender, and Steering Wheel Receptacles...............................................................................................................25 4-4 Picture of Auxiliary Brake Pedal.............................................................................25 4-5 Wiring Diagram for Data Acquisition System.........................................................27 4-6 Picture of National Instruments Si gnal Conditioning Carrier (SC-2345) and Modules....................................................................................................................27 4-7 Picture of Kistler Type 5210 Ca pacitive Accelerometer Power Supply..................33 4-8 Picture of PCB Model 480E09 Power Supply for ICP Laser Tachometer..............34 4-9 Wiring Diagram of Angular Rate Sensor Power Supply.........................................35 4-10 Picture of Equipment Installed in Trunk..................................................................40 4-11 Trunk Diagram Displaying the Approxi mate Location of Each Component..........40 4-12 Picture of Laser Tachometer, Mounting Bracket, and Drive Shaft..........................42 4-13 Picture of Installed Sting Potentiometer and Custom Pulley Assembly..................43

PAGE 10

x 4-14 Picture of Lateral Accelerometer, Longitudinal Accelerometer, and Angular Rate Sensor Installed Undernea th the Front Passenger Seat....................................44 4-15 Picture of Miniature Video Ca mera and Installation Bracket..................................46 4-16 Wiring Diagram of A ll System Components...........................................................47 4-17 Wiring Diagram Indicating the Locati on of All System Components in the Vehicle.....................................................................................................................48 5-1 Example Plot of a Typical Left Turn.......................................................................54 5-2 Plot of a Typical Normal Acceleration Test Maneuver...........................................55 5-3 Plot of a Typical Aggressi ve Acceleration Test Maneuver......................................56 5-4 Plot of a Typical Normal Deceleration Test Maneuver...........................................57 5-5 Plot of a Typical Aggressi ve Deceleration Test Maneuver......................................57 5-6 Plot of a Typical Normal Right Turn Test Maneuver..............................................58 5-7 Plot of a Typical Aggressi ve Right Turn Test Maneuver........................................59 5-8 Plot of a Typical Normal Left Turn Test Maneuver................................................61 5-9 Plot of a Typical Aggressi ve Left Turn Test Maneuver...........................................61 6-1 FHWA Road Course Generated Using GPS Coordinates with Intersections of Interest Highlighted and Labeled.............................................................................65 6-2 Example of Maneuver Plot Used to Veri fy That The Vehicle Dynamics Data and GPS Data Are Synchronized....................................................................................66 6-3 Plot of Maneuver 4B Generated Using stats.m With Markers Indicating the Start and End of The Turning Phase.................................................................................68 7-1 Picture of Right-Turn Channelization fr om FHWA Road Course Maneuver 2A....75

PAGE 11

xi Abstract of Thesis Presen ted to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Engineering VEHICLE DYNAMICS MEASUREMENT SYSTEM FOR EVALUATING OLDER DRIVER PERFORMANCE By Ethan J. Davis May 2006 Chair: Christopher Niezrecki Cochair: John Schueller Major Department: Mechanic al and Aerospace Engineering In 2003, The Occupational Therapy Department of The College of Public Health and Health Professions at The University of Florida founded the National Older Driver Research and Training Center (NODRTC). The NODRTC has received funding from the Centers for Disease Control (CDC), Federa l Highway Administration (FHWA), National Highway and Traffic Safety Administration (NHTSA), and the Florida Department of Transportation (FDOT). The NODRTC offers a ssessment of driving skills, referrals to health professionals to improve driving skil ls, and counseling in the use of alternative transportation for former drivers. This thesis discusses the development a nd implementation of a system to measure vehicle dynamics for the purpose of evaluating older driver performa nce. The system will be used for several research studies, although th is paper will only di scuss its use in an FHWA study. The goal of the study is to eval uate the effectiveness of intersection

PAGE 12

xii improvements recommended by the FH WA. The recommended intersection improvements are designed to make the inters ections easier to negotiate, particularly for older drivers. The vehicle dynamics measurement system consists of a data acquisition system, various sensors, and a GPS antenna. The sy stem is capable of measuring lateral acceleration, longitudinal acceleration, yaw (t urning rate), speed, steering wheel position, brake status, and turn signal status. There is also an event trigger that can be activated by an evaluator to mark the data so that the ev ent can be analyzed later. The GPS antenna is used to determine the location of the vehicl e while it is driving a pre-determined road course. The data are extracted for the intersectio ns of interest and an algorithm is used to divide the intersection into three distinct phases (approach, turn, a nd recovery). Several statistics are calculated for each variable a nd phase of the intersection. These statistics were analyzed using SPSS statistical analysis software. Overall the system was useful in acq uiring vehicle dynamics at particular intersections and inferring di fferences between old and young drivers. Furthermore, in relation to the FHWA study, th e results indicated that the recommended intersection improvements benefited both old and young driv ers and were more helpful to older drivers.

PAGE 13

1 CHAPTER 1 INTRODUCTION This thesis will discuss the development of a vehicle dynamics acquisition system to be used as an aid in driving researc h. The system will be in stalled in a 2005 Buick Century Sedan to be used by the National Ol der Driving Research and Training Center (NODRTC) to conduct research on older dr ivers. This chapter will discuss the background and motivation for older driver re search, specifically the research being conducted by the NODRTC. The goals and objectives of the vehicle dynamics acquisition system will also be presented. 1.1 Background The Occupational Therapy Department of th e College of Public Health and Health Professions at The University of Florida founded the Seniors Inst itute on Transportation and Communications in 2000. The name was la ter changed to the National Older Driver Research and Training Center to reflect it s natural focus after receiving funding in 2003 from the Centers for Disease Control and Prevention (CDC) and the Federal Highway Administration (FHWA). The NODRTC later received additional funding from the National Highway and Traffic Safety Ad ministration (NHTSA) and the Florida Department of Transportation (F DOT) and is currently conduc ting three research projects for the CDC, NHTSA, and FHWA. The purpose of the CDC research is to determine the effectiveness of particular screening tests in determining a personÂ’s driving abilities. The goal is to identify which tests can accurately predict a driver to be unfit to drive and prevent them from having to take a poten tially dangerous driving test. The NHTSA

PAGE 14

2 research is being performed to determine the validity and reliability of guidelines issued by the American Medical Associ ation (AMA) for diagnosing dr iving ability, specifically The PhysicianÂ’s Guide to Assessing and Couns eling Older Drivers. The FHWA research evaluates older driver perfor mance on roadway design features identified by the FHWA as possibly problematic for older drivers. The FHWA has issued guidelines for highway designs that are believed to incr ease the safe driving ability of older drivers. This research examines differences in driving performance be tween road designs that either meet or do not meet these guidelines. The NODRTC has two vehicles dedicated to driving research and evaluations. The design and implementation of the system describe d in this thesis may eventually be used for all of the research projects previously mentioned. Initially it will be used for the FHWA research and for that reason it will be the focus of the discussion. The system itself will be the same for the different studies. However, the specific goals for acquiring and processing the data will be unique. From th is point on, all discussion will be relevant to the FHWA research, particularly the data processing and analysis. 1.2 Older Driver Statistics The emphasis for evaluating older driverÂ’s driving abilities is motivated by the growing number of senior citizen s driving, their frequency of collisions, and the severity of their crashes. In 2003, there were 35.9 millio n people in United States age 65 or older. That number is expected to double by the ye ar 2030. When vehicle fatality rates are compared on the basis of miles traveled annually, the fatality rate for seniors is four times that of the 30-59 year old age group (DOT HS 809 328). An interesting note is that most of the older driver fatalities in 2000 occu rred during the daytime (81%), on weekdays (71%), and involved another vehicle (76%) ( DOT HS 809 328). All of these statistics

PAGE 15

3 suggest that older drivers are some of the most dangerous drivers on the road. Furthermore, with an inevitable increase in the number of older drivers from the baby boom generation, the number of collisions and fatalities as a result of older drivers is likely to increase. 1.3 Purpose of FHWA Research As previously mentioned, the FHWA resear ch project is focused on assessing older driver performance on roadway conditions identified by the FHWA as potentially problematic for older drivers. The FHWA has proposed certain roadway improvements that are intended to help all drivers negotiate an intersection. It is believed that these improvements will be most beneficial to older drivers. Two examples of these roadway improvements are an offset turning lane and a four foot wide receiving lane. The offset turning lane is the left-most lane when appr oaching an intersection. Instead of facing the oncoming turning lane, which is the standard for most intersections, the offset turning lane is positioned four or five feet furthe r left. This results in the driver having an improved view of oncoming traffic, thus maki ng it easier and safer to complete the turn. The four foot wide receiving lane is also in tended to make navigation of an intersection easier. The four foot wide receiving lane is an extension on the righ t side of the roadway after completing a left turn at an intersection. The widene d receiving lane allows the driver to make a wider turn and avoid having to turn sharply and possibly hit the median in the center of the road. The NODRTC designed a road course in Gainesville, Florida, to evaluate the effectiveness of these improved roadway safety features. The road course incorporates ten intersections. Five of the intersecti ons feature different recommended roadway improvements and the other five are their si milar counterparts without the improvements.

PAGE 16

4 They are all signalized intersec tions and all but one are left turn maneuvers. A street map of the FHWA road course showing the route is provided in Figure 1-1. Figure 1-1. Street Map of The FHWA Road Course in Gainseville, FL. A Driving Rehabilitation Specialist accompan ies the test subject as they drive the road course. The Driving Rehabilitation Specia list, also referred to as the evaluator, assesses the driverÂ’s performa nce during the designated inte rsections. The evaluator uses a scoring sheet that lists all of the possibl e driving errors for each intersection. As the

PAGE 17

5 driver navigates the course, the evaluator pa ys careful attention to each intersection and scores driving errors including lane mainte nance, forward vehicle gap acceptance, and overall stability. The FHWA research calls for at least 100 driver s age twenty-five or older to be evaluated. There are two hypothe ses being simultaneously tested in the FHWA study. The first is that a ll drivers will benefit from the improved safety features and will have fewer driving errors on th ese particular inte rsections. The second hypothesis is that older drivers will benefit from the new safe ty features even more than the rest of the population. The older drivers ar e expected to have mo re driving errors per person for the unimproved intersections, but perform similar to the younger drivers for the improved intersections. It is certainly possible that a ny age driver may exhibit poor driving performance for both the improved a nd unimproved intersections. This is why a statistical analysis will be performed on the behavioral da ta to detect any underlying trends. 1.4 Objectives In addition to the subjective evaluator scoring, the NODRTC wanted an objective way of measuring driver performance. The d ecision was made to design a system capable of measuring and recording vehicle dynamics such as acceleration and yaw. The system needed to be affordable and easy for an ev aluator to operate. The desired measurements were longitudinal acceleration, lateral acceleration, yaw, speed, brake status (on/off), and steering wheel position. Longitude and La titude provided by the Global Positioning System (GPS) were also needed to determine when the vehicle was at a particular intersection. The vehicle dynamics measuremen ts will ultimately be combined with the behavioral data to determine which kinematic measurements are the best indicators of

PAGE 18

6 driving performance. This will be performe d by a NODRTC research assistant once all of the subjects have been evaluated. Due to uncontrollable factors such as tra ffic and weather, each FHWA participant will encounter different driving scenarios. Th is will certainly make it more difficult to evaluate driver performance based on the vehicle dynamics. For example, if a subject approached an intersection to ma ke a left turn and had a gree n left-turn arrow, they would most likely slow down to make the turn and then accelerate up to a safe driving speed. On the other hand, if another subj ect approached the same inte rsection but had a red light, they would have to stop and wait for a green light to complete the turn. The longitudinal acceleration profiles for the two subjects would be drastically different. The first subject would most likely decelerate before the tur n, while the other subject has to accelerate after being stopped at the red li ght. Therefore, drivers will n eed to be compared based on the driving conditions that they encounter. In order to evalua te all of the drivers for the FHWA study on an equal playing field, th e NODRTC purchased a full-scale driving simulator. The simulator features a vehicle cab and three large projec tion screens for 180 viewing and a realistic driving experience. A picture of the simulator is provided in Figure 1-2. The ten intersections from the road course have been recreated for the simulator using the driving scenario programming langua ge. The simulator intersections were designed to be as similar to their road c ourse counterparts as was possible using the scenario programming language. The simulator has some limitations including only being able to produce perpendicular intersections. Si nce one of the road c ourse intersections is not perpendicular, it could not be recreated fo r the simulator. The advantage of evaluating

PAGE 19

7 drivers with the simulator is that they can all be scored for the exact same driving conditions. The simulator is capable of calcu lating all of the vehi cle dynamics that are being measured on the road course. Since all drivers will encounter the same scenarios, the simulator should be helpful in determ ining which measures are good indicators of driving performance. Figure 1-2. Picture of NODRTC Simulator sh owing vehicle cab and projection screens. The rest of this paper will focus on th e design, selection, installation, and data processing of the road course vehicle dynami c measuring system. A literature review of similar systems and driving research using such systems is provided in Chapter 2. The specific requirements for the system will be discussed in detail in Chapter 3. The overall design of the system and the selection and in stallation of components will be discussed in

PAGE 20

8 Chapter 4. System verification and sample driving maneuvers will be analyzed in Chapter 5. Chapter 6 will review the processing and analysis of the data as it relates to the FHWA study. The results and discussion will be pres ented in Chapter 7. Finally, a conclusion is provided in Chapter 8.

PAGE 21

9 CHAPTER 2 LITERATURE REVIEW A literature review was conducted to gather information for two purposes. The first was to investigate other research studies which in some way used vehicle dynamics and/or a quantitative assessment to evaluate driving performance. The second was to evaluate off-the-shelf or “tur n-key” devices which can be used to measure and record vehicle dynamics. It should be noted that th e NODRTC conducted an extensive literature review of more than 2,500 articles related to elderly driving, driv ing rehabilitation, and intersection roadway design. The literature revi ew presented in this chapter is solely concerned with instrumented vehicles, devices for acquiring vehicle dynamics, and relevant research studies. 2.1 Previous Research Studies Wood and Worringham (2005) used a quantita tive assessment to evaluate driving performance of drivers with Parkinson’s dise ase. An accredited professional driving instructor and a trained occupa tional therapist accompanied each driver as they navigated a 19.4 kilometer open road course. The occupa tional therapist “sco red seven aspects of driving performance at 147 locations along th e route (general obser vation, observation of blind spots, indication (signa ling), braking/acceleration, la ne position, gap selection, and approach” (177). The occurrence and type of error was recorded for each location. Also, each of the 147 locations “was allocated to on e of nine categories, including roundabouts, merging, pulling in/pulling out, traffic light and non-traffic light controlled intersections, reversing and parking, one way and two way traffic, and lane ch anging” (177). This

PAGE 22

10 allowed researchers to determine what types of errors were occurring most frequently and in which type of driving situation. The date wa s used to give each dr iver a driver safety rating. The study was used to correlate a driver ’s driver safety rating with the number of years since the diagnosis of Parkinson’s dise ase. Although this research did not make use of an instrumented vehicle, it showed how an objective quantitative assessment could be used to evaluate driving performance. Knodler et al. (2005) used a quantitative assessment and a “full-scale dynamic driving simulator as a tool for evaluating driver comprehe nsion by placing drivers in a fully interactive scenar io just as if they were actually driving” (270-271). The goal of the study was to determine which type of prot ected/permissive left-turn (PPLT) signal display was most effective. Again, this study di d not make use of an instrumented vehicle however it used a full scale simulator to objectively measure driver behavior. Another simulator study was conducted by Mc Gehee et al. (2001) to investigate how long it takes for a participant to adapt to the simulator. McGehee et al. objectively measured a participants adaptation by meas uring steering wheel reversals. “A steering wheel reversal is operationally defined as a deflection of the steering wheel away from a central or neutral position (i .e., where the wheels are completely straight) followed by a reversal of the steering wheel back to the neutral position” (2). In order to quantify steering wheel reversals a threshold of 6 degrees from neutral position was chosen. Steering wheel reversals were counted duri ng a 25-minute drive and were used to determine how quickly a driver could adapt to the simulator. The study showed that on average “older drivers need about three minutes to adap t and get the “feel” of the

PAGE 23

11 simulator. Before this time driving behavior in the simulator may not be representative of actual driving performance” (4). Rizzo et al. (2002) conducted a study using an instrumented vehicle because “road tests were designed to ensure that novice drivers know and can a pply the rules of the road, and not to test older, experienced driv ers who may be impaired” (291). Rizzo et al. also states “an instrumented vehicle pe rmits quantitative assessments of driver performance in the field, under actual road conditions. These measurements are not subject to the type of human bias that affect s inter-rater reliability on a standard road test” (291-292). The researchers developed an in strumented vehicle called ARGOS (the Automobile for Research in Ergonomics and Sa fety). “ARGOS was designed to examine objective indices of driving performance in normal and potentially unfit drivers and consists of a mid-seized vehi cle with extensive instrumenta tion and sensors hidden within its infrastructure” (292). ARGOS also utili zed video cameras and an example of the display output from ARGOS is shown in Figure 2-1. Figure 2-1. Display Output from ARGOS Instrumented Ve hicle (Rizzo et al., 292)

PAGE 24

12 The four video images were multiplexed onto a single display. The values for the various measurements were superimposed over the vi deo display. Uc et al (2004) used ARGOS to evaluate route-following and safety errors in drivers with early Alzheimer disease and also provided a more detailed description of ARGOS. “Experimental performance data (e.g.,, steering wheel position, normalized acceler ation and vehicle speed) were digitized at 10 Hz and reduced to mean SD (standard deviation), or count. Dr iver’s lane tracking and visual scanning activity of the envi ronment were recorded by videotape at 30 frames/s using miniature lipstick-sized cameras mounted unobtr usively within the vehicle” (833). It is evident from Figure 2-1 that ARGOS has fo rward, rear, driver’s face, and driver’s hands cameras. Furthe rmore, ARGOS was capable of measuring accelerator (gas) pedal and brake pedal depr ession (measured as a percent of fully depressed). ARGOS was used successfully to quantitatively assess drivers with Alzheimer disease while performing a specific driving task. Dingus and Hulse (1997) used an instrument ed vehicle to determine the effects of age and experience on navigation while driving. “The system consisted of video cameras to record pertinent event and eye movement da ta, an experimenter control panel to record incident classification data, and sensors fo r the detection of variations in driving performance and behavior” (185). The system utilized four video cameras with nearly identical views as those used with ARGOS. “T he system provided the capability to store a line of numerical data every 0.1 seconds” ( 185). It utilized accelerometers to measure lateral and longitudinal acceler ation forces. The system also measured steering wheel position, vehicle speed, and brak e pedal status (on/off). The research focused on using lateral acceleration measurements to indicate dr iver awareness. It also made use of the

PAGE 25

13 video cameras to time the driver’s gl ancing eye movements. Although the study was mainly concerned with driver navigation and aw areness rather than dr iver performance, it was able to use an instrumented vehicle to objectively quantify driver behavior. Boyce and Geller (2002) used a vehicle with concealed video cameras and transducers to evaluate driver behavior specifically younger males. The “computer generated dependent measures, in concert with real-time c overt video recording of the participants’ driving, allowed for unpreceden ted opportunities to perform a behavior analysis of driving performance in the context of normal driving. Thus, problems associated with truthfulness of self-report, in frequency of occurrence, and error variation due to in-vehicle observers were minimized or avoided altogether” (52). In addition to video recording, the vehicle also measured and recorded driver safety-belt use, number of times a turn signal was used, “vehicle veloc ity including average speed, velocity changes, and velocity variance, and fo llowing distance measured in meters” (53). Again, the video camera configuration was similar to that f ound in ARGOS. Although not directly related to driver performance, this study made use of an instrumented vehicle to measure driver behavior in the absence of a driving evaluator. 2.2 Devices This section will offer a brief review of several commercially available products that are used to measure vehicle dynamics. These devices are calle d turn-key devices because they can be installed and ready to us e right out of the box. In general, they are designed to serve a single purpose and us ually not expandable or configurable. 2.2.1 DriveCam The DriveCam is a device that is primar ily used for vehicle fleet tracking and maintenance. It is installed behind the central rear view mirror in a vehicle. It features

PAGE 26

14 two video cameras to record forward-facing a nd rear-facing views. In addition, it has a longitudinal and a lateral accelerometer. A pictur e of the DriveCam is provided in Figure 2-2. Figure 2-2. Picture of DriveCam. The DriveCam uses removable digital me dia and as a result can only store small amounts of data (about 20 seconds). It records continuously and in th e case of an event (e.g., hard turn or braking), it will store 10 seconds before an d 10 seconds after the event. The measurements from the accelerometers ar e displayed at the bottom of the screen during video playback. The accelerometers are only accurate to a tenth of a g The device can also be triggered by the user to save a sp ecific event. Due to the limited number of measurements and lack of expandability, in addition to the short recording time, the DriveCam was not a suitable solution for the FHWA research. 2.2.2 Vericom Vericom Computers Inc. manufactures th e VC3000 which is an on board vehicle dynamometer. It features a two axis acceleromet er which is used to measure longitudinal and lateral acceleration. It inte grates the longitudinal acceler ation to calculate speed and

PAGE 27

15 distance traveled. It is a self contained unit a nd is usually installed with suction cups to the inside of the passenger side windsh ield. The VC3000 can accept up to 6 additional sensor inputs. These sensors are sold by Vericom and only a limited selection is available. The data collect ed with the VC3000 can be downloaded to a computer and analyzed using the provided software. Fi gure 2-3 shows a picture of the Vericom VC3000. Figure 2-3. Picture of Vericom VC3000. The VC3000 is primarily used for vehi cle performance tests (e.g., horsepower, braking) and traffic accident investigation and re construction. These tests are usually of short duration and consequently the VC3000 was designed with only enough flash memory to store approximately 21.8 minutes of data. This was the main reason that the VC3000 could not be used for the FHWA resear ch. Furthermore, it is not completely expandable and the software has limited capabilities.

PAGE 28

16 2.2.3 Dewetron Dewetron is a worldwide manufacturer of portable data acquisition systems. They offer several different products depending on the requirements of the application. Their products are essentially persona l computers combined with data acquisition hardware and DewetronÂ’s own proprietary software. The units are capable of measuring nearly any type of signal including video and GPS. Dewetron o ffers desktop as well as portable products including some that are ba ttery powered. Figure 2-4 show s a picture of one of the Dewetron units. Figure 2-4. Picture of Dewetron DEWE5000. The Dewetron systems are very sophisticat ed and the software is specifically designed for evaluating vehicle performance. However, in the context of the FHWA research, the Dewetron system would only be us ed as a data recorder. The software used to analyze the data from the FHWA road cour se would still have to be developed. As a

PAGE 29

17 result, the Dewetron system would be too expens ive to use especially considering that its full functionality is not being utilized.

PAGE 30

18 CHAPTER 3 SYSTEM REQUIREMENTS After reviewing several off-the-shelf pr oducts with their own advantages and disadvantages, a decision was ma de to design a system from sc ratch that satisfies all of the requirements. The system must be capable of measuring accelerations, speed, yaw, steering wheel position, longitude, latitude, a nd brake and turn signal status. Additionally, the system should be affordable, easy to ope rate, and flexible. Designing a customized system allows it to be confi gurable and used for a wide va riety of applications. This chapter will discuss the specific goals and performance requirements of the system as well as additional considerations including cost, ease of use, and flexibility. 3.1 Performance The system should offer good performance in the form of accuracy and high sampling rate. This system will likely be used for future studies other than the FHWA research. Therefore, it is necessary that the system of fers a high level of performance such that future upgrades will not be necessary. The syst em should have a resolution of at least 16bits to provide accurate sampling of each cha nnel. A 16-bit system would allow a voltage resolution of 0.0001526 volts for an input range of 10 volts. In addition to accuracy, sampling rate is another important performan ce specification. A digital data acquisition system can only sample an analog signal at discrete times. Therefore, a high sampling rate is needed to accurately reproduce th e original analog signal. If frequency measurements are desired, the sampling freque ncy must be at leas t twice the highest frequency component of the analog signal accord ing to the Nyquist Theorem. This is to

PAGE 31

19 prevent aliasing, which occurs when the sa mpled signal appears to have a different frequency than the original analog sign al. Frequency measurements, for example frequency of steering wheel reversals, ar e not being considered for the FHWA study. However, it will still be necessa ry to accurately recreate the original signal. Furthermore, frequency measurements (e.g., steering wheel input) may be utilized in future NODRTC research. It was decided that the system should be capable of sampling each channel at a minimum of 10,000 samples per second (10k Hz). This will ensure accurate reconstruction of the analog si gnal and allow for frequency measurements up to 5000 Hz. This is certainly capable of measuring a ny input physically gene rated by a human. Also, it is doubtful that any of the vehicle dynamics will have frequency components of this magnitude. 3.2 Transducers The accelerometers used for measuring longitu dinal and lateral acceleration need to be able to accurately measure at low freque ncies (i.e., less than 10 Hz). Measurement of the high frequency vehicle vibr ation is not important to this study. The changes in acceleration will be gradual and the maximum acceleration in any direction under normal driving conditions is not expected to exceed 1 g. The transducer that is used to measur e the steering wheel position should be nonintrusive. It should be able to accurately provide the positi on of the steering wheel at any instant. The sensor should not add any extra resistance to the steering wheel that may affect the driverÂ’s ability to turn the wheel. As with the accelerometers, the transducer that measures yaw should be able to measure accurately at low frequencies. U nder normal driving conditions, the maximum yaw is not expected to exceed 90/sec. This would be the equivalent of completing a 90

PAGE 32

20 turn in one second, which is unrealistic for everyday driv ing. The yaw sensor should be able to determine the direction and rate that the ve hicle is turning. The transducer used to measure vehicle speed should be non-intrusive. Velocity can be derived from the longitudinal acceleration, but a more direct measurement would be preferable. It may be possible to ac quire vehicle speed via the carÂ’s on board diagnostic system. The data acquisition system should be able to determine if the driver is using the brake and the turn signals. Add itionally, the evaluators requested that an event trigger be installed in the vehicle cab. The event trigger will be used as a way to mark the data in the case of an adverse event for later review. 3.3 GPS A global positioning system receiver and antenna is needed to acquire the vehicleÂ’s location while driving on th e road course. The receiver should be WAAS capable and easily integrated into the data acquisition system. The Wide Area Augmentation System (WAAS) consists of 25 ground reference stations and two geostationary satellites orbiting above North America. The ground reference stations monitor the GPS satellites and correct for errors caused by satellite position error, clock drift, a nd atmospheric effects (Garmin Ltd.). The corrected signals are then relayed to the geostationary satellites. The correction message can be received from the geostationary satellites using a standard WAAS-capable GPS receiver. A GPS receiver using WAAS has an accuracy of less than 3 meters 95 percent of the time. Differential GPS (DGPS) uses a local reference receiver, whose location is known exactly, to calculat e the error in the G PS signals. The error signals are then broadcast, usually via radi o, to the other GPS rece ivers. DGPS requires a

PAGE 33

21 reference receiver and radio transmitters and is consequently more costly than a WAAScapable receiver. As a result, DGPS will not be considered for this project. 3.4 Video The video system should consist of a forw ard view, rear view, and a view of the driver. The driver view will be used after a driving assessment to evaluate the driverÂ’s eye movements. All of the views should be recorded simultaneously to a single digital medium. Digital video will provide distortion-free frame by frame playback. The media needs to be long enough so that the entire ro ad course will fit on a single recording. 3.5 Ease of Use The data acquisition system will be primarily operated by Occupational Therapists and therefore the system should be user frie ndly and able to a ccommodate users with minimal technical experience. The end user shou ld not have to calibra te or configure the system in any way. Consideration will be take n in the design of the system such that the user can operate it after receiving minimal training. 3.6 Cost The NODRTC has received substantial fundi ng from the FHWA. In addition to the data acquisition system, the budget must be used for purchasing the ro ad course vehicle, driving simulator equipment, and vehicle cab for the driving simulator. Therefore the costs for the vehicle dynamics acquisition sy stem should be minimized if possible. Regardless of the data acquisition system use d, the transducers and sensors will cost the same. The system should be expandable so th at the option to use additional sensors for the FHWA or other rese arch is available.

PAGE 34

22 3.7 Flexibility This system will likely be used for other older driver research at the conclusion of the FHWA study. The system must be flexible and adaptable to other areas of research. For example, the FHWA research is only concerned with 10 specific intersections whereas other research may be interested in driver performance on an entire course. Furthermore, different measurements, possibl y physiological, may be desired for future research.

PAGE 35

23 CHAPTER 4 SYSTEM DESIGN This chapter will discuss the selection, installation, and operation of the entire system including video. The selection of components including the data acquisition system, transducers, GPS, and video equipment will be discussed in detail. A description of the vehicle and its driving aids will also be presented. 4.1 Vehicle Description The system was installed in a 2005 Buick Century Sedan. A picture of the vehicle is provided in Figure 4-1. Figure 4-1. Exterior View of 2005 Buick Ce ntury Sedan NODRTC Research Vehicle. The vehicle is equipped with adaptive driving equipment for handicapped drivers. There is a left foot accelerator pedal adapter that mounts to the floor and replaces the standard accelerator pedal. The adapter is secured to a bracket that is permanently attached to the floor using locking pins. This adaptation is for drivers who have limited or

PAGE 36

24 no use of their right leg. The adapter features a metal plate to prevent the driver from accidentally pressing the original accelerato r with their right leg. Figure 4-2 shows a picture of the left foot accelerator adapter. Figure 4-2. Picture of Left Foot Accelerator Pedal Adapter. For driverÂ’s who have limited use of bot h legs, there are hand controls. The hand control is a lever on the left si de of the steering wheel that is mechanically linked to the brake and accelerator pedals. The driver ro tates the lever to accelerate and pushes it forward to brake. Since the driver must use their left hand to control the accelerator and brake, there is a turn signal extender that allo ws the driver to activate the turn signal with their right hand. The car also features three receptacles on the steering wheel for spinner knobs. The knobs enable a driver suffering from ar thritis to turn the steering wheel easier. A picture of the hand controls and steering wheel receptacles is shown in Figure 4-3.

PAGE 37

25 Figure 4-3. Picture of Hand Controls, Turn Signal Extender, and Steering Wheel Receptacles. In addition to the adaptive dr iving equipment, the car is al so equipped with an auxiliary brake pedal for the evaluator. The auxili ary brake pedal is shown in Figure 4-4. Figure 4-4. Picture of Auxiliary Brake Pedal.

PAGE 38

26 The auxiliary brake pedal is mechanically linke d to the standard brake pedal using a steel cable and pulleys. The auxiliar y brake is used by the evaluator if a driver exhibits reckless driving behavior. Also, the car was pur chased with a bench seat which allows the evaluator to grab the steeri ng wheel and take control of the vehicle if necessary. 4.2 Selection of Components 4.2.1 Data Acquisition System The data acquisition system is the centerp iece of the design. All of the sensors are connected to it and it is used to record all data. It is also the user interface for the entire system. The system was installed in a car a nd must be mobile, so a desktop computer would not be practical. The decision was ma de to use a powerful laptop mated with a data acquisition card. The laptop chosen is a Dell Inspiron with a Pentium 4 2.4 GHz processor and 512 MB of RAM. The data acq uisition card is a National Instruments DAQCard-6036E with PCMCIA connectivity. The card is 16-bit a nd is capable of 200,000 samples per second. This allows samplin g rates of 10,000 samples per second for 20 channels. The card is connected to a Si gnal Conditioning Carrier (SCC) via a 68-pin cable. The SCC is the link between the sensor s and the data acquisi tion card. Each SCC can hold several signal conditioning modules, which are customized inputs depending on the type of sensors used. The modules allo w connection of different transducers (e.g., accelerometers, thermocouples, and strain gauge s) to a single measur ing device. The SCC most appropriate for this application is the SC-2345. The SC-2345 can hold up to 20 modules (16 inputs, 2 outputs, 2 counters) a nd has 8 digital I/O lines built in. There are three options for powering the SC-2345: 5 volt DC power from a DAQ device, AC power, or a 7 to 42 volt DC power supply. AC pow er is required to power the laptop, so it will also be used to power the SC-2345. In addition to selecting the right module for each

PAGE 39

27 sensor input, the physical conn ection can also be specified. There are a wide variety of connections including BNC, 9-pin Dsub, SMB, MIL-Spec, and LEMO. These connections are referred to as panelettes and any combination may be used. For this application, all of the panelettes are BN C and have two inputs per panelette. The panelettes have lead wires which are connect ed to the screw terminals of each module. This allows each sensor to be connected using its native connect or. A simple wiring diagram is shown in Figure 4-5 and a pict ure of the SCC is provided in Figure 4-6. Figure 4-5. Wiring Diagram for Data Acquisition System. Figure 4-6. Picture of Nationa l Instruments Signal Conditi oning Carrier (SC-2345) and Modules. The selection of modules will be discussed fo llowing the description of the sensors. The sensors must be selected firs t so that the proper type and input range of each module can be chosen.

PAGE 40

28 4.2.2 Transducers The section will discuss the selection of each transducer. The basis for selection and relevant specifications will be provided. Accelerometers It was determined that the accelerometers used to measure longitudinal and lateral acceler ation need to be capable of measuring at extremely low frequencies. The decision was made to use a pair of static acceler ometers manufactured by Kistler. These accelerometers can measur e accelerations in the range of 0 to 300 Hertz. The accelerometers utilize a variable capacitance sensing element. A small inertial mass is attached to a flexure element cantilevered between two plates. The mass deflects under acceleration resulting in a change in the capacitan ce between the two plates. Internal circuitry provides AC excitati on and synchronous amplitude demodulation resulting in a voltage output that is propor tional to the applied accel eration. There are two models available (8310A2 and 8310A10) w ith ranges of either 2 or 10 g Â’s. The 10 g model was chosen to measure long itudinal acceleration while the 2 g model was chosen for lateral acceleration. The 2 g model is used to measure lateral acceleration because even a high-end sports car cannot exceed 1 g laterally on a skidpad (Road & Track). By choosing the accelerometer with the smalle r range, the lateral acceleration can be measured more precisely. The 10 g model was chosen to measure longitudinal acceleration. Under normal driving conditi ons the longitudinal acceleration is not expected to exceed 2 g Â’s in either direction. However, it may be possible to exceed 2 g Â’s under heavy breaking or in a collision. Therefore the 10 g accelerometer is used to measure longitudinal acceleration to pr event saturation of the signal at 2 g Â’s. Technical specifications for the accelerometers and all other transducers are provided in Appendix

PAGE 41

29 B. Installation and power requirements for th e accelerometers will be discussed later in the chapter. Angular rate sensor (Yaw) An angular rate sensor was chosen to measure the vehicle yaw (i.e., rate of rotation). The a ngular rate sensor is manufactured by BEI Technologies Inc., and is called the Gy roChip (AQRS-00075-111). It is a compact, rugged micromachined vibrati ng quartz gyroscope, designe d specifically for demanding automotive and commercial applications (BEI 964172 Rev. A). The GyroChip uses a double-ended quart tuning fork made from monocrystalline piezoelectric quartz. As a result of the Coriolis effect, there is a volta ge output directly propor tional to the rate of rotation. The use of a single piezoelectric crysta l makes it less susceptible to variations in temperature and extends its life. The maximum rate of rotation the GyroChip is capable of measuring is 75 degrees per s econd. It is expected that this rate woul d not be exceeded under normal driving conditions. If the car were to exceed this rate and sustain it, that would mean it is capable of completing a 90 degree turn in just over a second. This is certainly not realistic, howe ver a strong jerk on the steering wheel may cause the car to approach the maximum rate. Furthermore, th e GyroChip is designed especially for use with automotive stability control systems so it is an appropriate choice for measuring vehicle yaw. String potentiometer (s teering wheel position) Measuring the position of the steering wheel accurately and in a non-obtru sive manner proved to be difficult. An absolute rotary encoder was first considere d, but the decision was made to use a string potentiometer and custom made pulley. It is not possible to mount a ro tary encoder on the steering shaft so it would have to be connect ed using a pulley or gear. Knowing that a

PAGE 42

30 pulley or gear would be needed to connect a sensor to the shaft, a string potentiometer was chosen instead. Most absolute rotary encoders transmit a digital word to the data acquisition system. It is much easier to use a string potentiometer because it outputs a voltage directly proportional to displacement. When attached to a pulley on the steering shaft, it could accurately measure the pos ition of the steering wheel. A PT1DC CableExtension Position Transducer (PT1DC50-BK-Z10-MC4-SG) produced by Celesco was selected. Several aspects of the PT1DC ar e customizable includi ng cable length, cable exit location, output voltage, and connection type. The PT1DC features internal power conditioning so it does not require a regulated power supply. The cable tension is only 5 ounces which will apply a negligib le amount of torque (less th an 0.1 ft-lb.) to the steering shaft. This small amount of torque will have negligible effect on the position of the steering wheel or the effort needed to rotate it in either direction. The maximum available range of 50 inches was selected in case th e string potentiometer could not be mounted close to the steering shaft. If absolutely necessary, several pulleys could be used to route the cable from the sensor to the pulley on the steering shaf t. The design of the custom made pulley for the steering shaft and the inst allation of the string pot entiometer will be discussed later in the chapter. Laser tachometer (Speed) As with steering wheel position, measuring speed in a non-obtrusive manner proved to be a challe nge. Obtaining vehicle speed from the OnBoard Diagnostic system was investigated but proved to be too difficult to integrate with the data acquisition system. The LaserTach ICP tachometer from The Modal Shop Inc. was chosen to measure vehicle speed. The LaserTach emits a laser and has a sensor to detect when it is reflected. A re troreflective target is placed on the rotating shaft to reflect

PAGE 43

31 the laser. The LaserTach produces a voltage pulse each time the laser is reflected. The frequency of the pulse signal is the rotational frequency of the shaft. Integrated Charge Pre-Amplifier (ICP) sensors have a built-in mi croelectronic amplifier which results in a low impedance coaxial system. An external ICP power supply or a data acquisition system capable of powering ICP sensors must be used to provide power to the sensor. Event trigger, brake, and turn signals The event trigger, brake, and turn signals are essentially identical They are simply coaxial cables connected to each signal individually. The coaxial cables are wired in pa rallel with the brake a nd turn signal lights. The other end of each coaxial cable is connect ed to the data acquisition system. The event trigger is slightly different in that it must be supplied a voltage. The event trigger is a simple push-button with one terminal connected to +12 volts from the vehicleÂ’s electrical system and the other terminal wired to th e data acquisition system. The event trigger input to the data acquisition system is coupl ed with a ground wire to determine when the trigger is pushed. All of these signals will generate approximately 12 volts when activated. A simple logic statement can be impl emented in the software such that if the voltage is above a certain threshold the signal will be considered activated. 4.2.3 Modules As previously mentioned, the input modules for the SCC can be selected to match the output of each sensor. Once all of the tr ansducers were selected, the appropriate modules could be chosen. All of the sensors chosen will require modules with analog inputs. Each analog input module contains tw o individual inputs and are available with different input voltage ranges. A total of five modules will be needed to accommodate all of the inputs. It was determined that (2) 5 volt, (1) 10 volt, and (2) 20 volt modules would be needed. The specifications for each module are provided in Table 4-1.

PAGE 44

32 Table 4-1. Technical Specifications fo r Signal Conditioning Analog Input Modules. Model Quantity Channels Input Range Bandwidth SCC-AI02 2 2 20 volts 10 kHz SCC-AI03 1 2 10 volts 10 kHz SCC-AI04 2 2 5 volts 10 kHz Each channel of each analog input module c ontains an instrument ation amplifier, a lowpass filter, and a potentiometer for calibrati on. Table 4-1 displays the output range for each transducer and the input range for the selected module. Table 4-2. Transducer Output Ranges and Module Input Ranges. Transducer Output Range (V) Module Input Range (V) Longitudinal Accelerometer 0 – 2 SCC-AI04 5 Lateral Accelerometer 0 – 2 SCC-AI04 5 Angular Rate Sensor 0.25 – 4.75 SCC-AI04 5 String Potentiometer 0 – 10 SCC-AI03 10 Laser Tachometer 0 – 1 SCC-AI04 5 Brake 0 – 12 SCC-AI02 20 Left Turn Signal 0 – 12 SCC-AI02 20 Right Turn Signal 0 – 12 SCC-AI02 20 Event Trigger 0 – 12 SCC-AI02 20 All of the modules will be connected to pa nelettes with BNC connections. BNC offers a quick and reliable connection a nd is also the native form of connectivity for all of the sensors. 4.2.4 Power Supplies The accelerometers, angular rate sensor and laser tachometer all require specialized power supplies. The string potenti ometer power supply doe s not need to be regulated, nor does the supply voltage for the event trigger. The brake and turn signal connections will be powered by the voltage provided to the respective lights. The power supplies for the accelerometers ar e also manufactured by Kistler and run off of a single 9-volt battery. The power supplies are specific ally designed for capacitive accelerometers. Each power supply is conn ected to an accelerometer via a 4-pin Microtech connector. The output of each pow er supply uses a BNC connector and is

PAGE 45

33 easily connected to the data acquisition system. Each power supply provides the appropriate voltage and current to each accelero meter. They feature internal jumpers to adjust the gain and a potentiometer to adjust the output voltage bias The model is Type 5210 and a picture is provided in Figure 4-7. Figure 4-7. Picture of Kistler Type 5 210 Capacitive Accelerometer Power Supply. The laser tachometer is an ICP sensor so a standard off-the-shelf ICP power supply may be used. The selected power supply is a Model 480E09 manufactured by PCB. The power supply uses BNC connections for both the input and output. It features an adjustable gain and provides the proper volta ge and current for the laser tachometer. The power supply operates off of three 9-volt ba tteries providing approximately 50 hours of battery life. A picture of the power supply is provided in Figure 4-8.

PAGE 46

34 Figure 4-8. Picture of PCB Model 480E09 Po wer Supply for ICP Laser Tachometer. A suitable power supply for the angular rate sensor could not be found so it had to be fabricated. The angular rate sensor requi res a regulated power supply of 5 volts DC. The decision was made to build the power supply using a voltage regulator integrated circuit. Although the string potentiometer doe s not require a regulated power supply, it was incorporated into the angular rate power supply so that both could be controlled using a single DPST (double pole, single th row) switch. This will reduce the number of switches that need to be switched when usi ng the system. A wiring diagram of the power supply, including the DPST switch, is provided in Figure 4-9.

PAGE 47

35 Figure 4-9. Wiring Diagram of Angu lar Rate Sensor Power Supply. The power supply accepts +12 volts to provide power to the regulat ing circuit and + 24 volts to power the string potentiometer. A LM7805 voltage regulator chip is used to provide a constant + 5 volts to the angular rate sensor. Th e 1000 F capacitor helps to maintain constant input to the regulator while the 0.1 F capacitor eliminates any high frequency pulses that may in terfere with operati on of the regulator. The LM7805 requires an input voltage at least two to three volts greater than the output voltage. The carÂ’s electrical system provides approximately + 12 volts to the regulator chip, which satisfies the input voltage criterion. The + 24 volts is provided by a DC-DC converter normally used to supply power for a laptop. The DC-D C converter accepts + 12 volts and steps it up to + 24 volts. The + 24 volts does not need to be regulated so a simple switch can be used to turn it on and off. The LED is used to indicate the on/off status of the power supply. Since the power supply receives power from the carÂ’s battery, it could drain the

PAGE 48

36 battery if left on. A 3-pin connector is used to connect to each sensor to the power supply. BNC connections are used for the outputs of the power supply for a quick a reliable connection to the data acquisition system. A ll of the circuitry for the power supply is contained in a plastic enclosure. The inputs to the power supply use 2-pin connectors so the entire power supply can be removed or replaced quickly. AC power is provided by a Vector Maxx 750-watt power inverter. The power inverter is powered by the carÂ’s 12 volt electri cal system. The output is a modified sine wave with a frequency of 60 Hz and an amplitude of 115 volts (rms), similar to a standard household outlet. The power invert er is connected to a Monster Cable brand power strip for distribution of the AC power. All of the devices that use AC power are connected to the power strip. The power inve rter features shortcircuit and overload protection circuitry. 4.2.5 GPS A Garmin GPS 16 was selected for the GPS receiver. The GPS 16 is WAAS capable and uses the NMEA-0183 standard. It is enclosed in a rugged, waterproof enclosure and features a magnetic bottom for quick installation. Since the laptop does not have a serial port, a USB/Seri al converter will be used to connect the receiver. The converter also provides power to the receiver which cannot be done with a standard serial port. 4.2.6 Video The video system consists of four ultra-mini pinhole cameras (CCM630P), a quad multiplexer (QC800), and a portable mini -DV recorder (Sony GVD-1000). The cameras are color video cameras and measure 1.25 inches square. There is a choice of either a 78 or 90 viewing angle. The version with th e 90 viewing angle was chosen because it

PAGE 49

37 provided greater visibility for each camera. Each camera uses a 1/3-inch high-resolution CCD sensor and features an automatic iris and backlight compensation. The cameras are powered with 12 volts DC so the carÂ’s electrical system will be used to power them. A fuse was installed for protection as well as a switch to turn the cameras off when not in operation. The quad multiplexer is a device most commonl y used for security applications. It combines the video signal from up to four cameras into one signal. There are several options for displaying the four views: a large main view with the ot her three views shown at the bottom, a large main view with a sm aller window for a second view, a large main view sequentially shifting between all cameras and all four views shown at the same time in equally sized windows. The latter opti on was selected since e ach view is equally important. A portable mini-DV record er (Sony GV-D1000) was chos en for its exceptional quality and ease of use. The recorder features a 4-inch LCD screen to verify the system is working properly and cameras are aligne d properly before recording. The mini-DV format is a digital format with up to 500 lin es of horizontal reso lution. This is better resolution than Hi-8 and S-VHS and it is digital. The digital format allows it to be transferred to a computer easily for bac kup or editing. Some mini-DV tapes feature memory chips that can be used to save info rmation such as the date and time of the recording. The memory chip can also be used to build a table of contents. The table of contents contains a list of specific points in the recording. These points can be used for easier access during playback or fo r editing to be done later.

PAGE 50

38 The Quad multiplexer and mini-DV record er use standard AC power and are powered by the power inverter. Composite (RCA ) video cables are used to connect the cameras to the multiplexer and the multiplexer to the recorder. The mini-DV tapes can record 90 minutes in EP mode. This is long e nough to record the enti re road course on a single tape since the road course takes appr oximately 1 hour and 20 minutes to complete. The entire video system is completely inde pendent from the data acquisition system. The only device the two systems share is the power inverter. The video will be used to further review a subjectÂ’s performance if necessary. It may also be used as a training tool for new evaluators. 4.2.7 Software National InstrumentsÂ’ proprietary software LabVIEW (version 6.1) will be used to operate the data acquisition system. LabV IEW is a graphical programming language designed specifically for interaction with Natio nal Instruments hardware. It features builtin measurement and control subroutines, called virtual instruments, which allow for fast and easy program development. A combination of several virtual instruments are wired together to create a working program. A virt ual instrument can be changed or created from scratch to perform a custom operation. Th e user interface for the final program can be configured to hide or show various controls (inputs) and indicators (outputs). The program to be used for this system is fairly simple. It will be requi red to measure all of the channels from the data acquisition system a nd record the data to the laptop hard drive. National Instruments also provides the M easurement and Automatic Explorer (MAX) which is used to configure the entire system. The National Instruments hardware including the DAQ-Card, SCC, and SCC modules are entered in to MAX. Additionally,

PAGE 51

39 MAX can be used to monitor each channel to ensure that each sensor is working properly. GPSLog software developed by U.S. Positioning LLC will be used to record the GPS data. GPSLog is a simple program that records the output sentences from any NMEA-0183 compliant GPS receiver. NMEA -0183 is a standard defining signal requirements, data transmissi on protocols, and sentence form ats for 4800-baud serial data buses, which are widely used for GPS receiver communication (National Marine Electronics Association). GPSLog can save either the raw data sentences or just important information such as long itude, latitude, speed, and heading. The MathWorks MATLAB software will be used to process and analyze the data. MATLAB is a high-level technical computing language which is computationally faster than traditional programming languages lik e C++ and Fortran (MathWorks Inc.). MATLAB is the ideal applica tion for inputting and analyzin g the data. It features toolboxes which contain subroutines for pe rforming data analysis and statistics. MATLAB is a very powerful application howev er it will be mainly used for simple operations such as matrix manipula tion and statistical analysis. 4.3 Installation The system was installed in the vehicle in the least conspicuous way possible in order to have a minimal impact on the driver. The driver is most likely driving a vehicle different from their own and any equipment in plain view may make them nervous. The majority of the equipment including the power inverter, SCC, power supplies, and video equipment was installed in the trunk of the vehicle. A sheet of medium-density fiberboard (MDF) was cut to fit in the trunk and was used as a mounting platform for the equipment.

PAGE 52

40 A picture of the equipment installed in th e trunk and a diagram showing the approximate location of each component is shown in Figures 4-10 and 4-10 respectively. Figure 4-10. Picture of Equipment Installed in Trunk. Figure 4-11. Trunk Diagram Displaying the A pproximate Location of Each Component. (QUAD Video Multiplexer, PS Power Supply, ARS Angular Rate Sensor, SP String Potentiometer, Acc. Accelerometer)

PAGE 53

41 The laptop was originally installed in th e trunk however the Occupational Therapists requested it be located in th e back seat so that it could be operated from inside the vehicle. There were few options for the locations of th e laser tachometer and string potentiometer. The laser tachometer had to be installed near either one of the front wheels because the vehicle is front-wheel driven and does not have a rear axle. The string potentiometer needed to be installed close to the steering sh aft because it must be mechanically attached to the shaft. A bracket included with the la ser tachometer was fastened to an existing mounting bracket for one of the carÂ’s brake li nes. The laser tachometer is threaded and was installed in the bracket using the provide d hardware. The front axle was painted flat black and a -inch wide piece of retroreflective tape was inst alled. A picture of the laser tachometer setup is provided in Figure 4-12. The tachometer was aimed at the retror eflective tape and the laser beam was focused using the adjustment screw. When th e sensor is powered, a LED illuminates to indicate that the laser is bei ng reflected and measured properly. The string potentiometer is physically a ttached to the steering shaft via a pulley and therefore had to be installed relatively close to the shaft. It was not possible to remove the steering shaft in order to instal l an off-the-shelf pulley. The only feasible option was to design a pulley that could be att ached to the shaft wit hout having to remove it. The pulley is split in half so that it can be secured to th e shaft using socket cap screws. It features a groove to guid e the cable from the string pot entiometer as it winds around the pulley. The steering shaft is not perfectly round and consequently the inside of the pulley had a unique geometry. The pulley wa s machined out of a scrap piece of

PAGE 54

42 aluminum using an end-mill and lathe. The best location that for the string potentiometer that allowed the cable to reach the pulle y unimpeded was on the lower part of the firewall. The string potentiometer was secured to the firewall using J-B Weld two-part epoxy. A picture of the string potentiometer a nd pulley assembly installed in the vehicle is provided in Figure 4-13. Figure 4-12. Picture of Laser Tachomet er, Mounting Bracket, and Drive Shaft.

PAGE 55

43 Figure 4-13. Picture of Inst alled Sting Potentiometer a nd Custom Pulley Assembly. A drawing of the custom pulley was generate d using Pro Engineer modeling software and can be found in Appendix C. The accelerometers and angular rate sensor do not need to be located near any particular area of the car. The main requirement is that they are attached securely to a rigid part of the vehicle in order to measure accelerations and yaw accurately. Furthermore, the three sensors should be locate d as close to the center of the car as is feasibly possible. This is to reduce the am ount of measured acceleration that may be a result of suspension movement. The best locatio n that would also allow for the sensors to be out of sight was under either one of the fr ont seats. The driverÂ’s seat features power controls and as a result there are several mo tors underneath it. Ther efore, the only option was to install the sensors under the front passe nger seat. A piece of carpet and insulating

PAGE 56

44 foam was removed to allow access to the struct ural members of the vehicle. A frame rail that ran laterally across the vehicle was us ed as a guide to position the longitudinal accelerometer. The longitudinal accelerometer wa s located along the frame rail such that it measured in a direction perp endicular to the frame rail. The orientation of the angular rate sensor did not matter as long as it wa s positioned level so that it measured rotation about the vertical axis of the car. It was positioned along the frame rail so that a perpendicular face on the angular rate sensor could be used as a mounting location for the lateral accelerometer. A ll of the sensors were attached using J-B Weld two-part epoxy. A picture of the sensors installed beneath th e passenger seat is provided in Figure 4-14. Figure 4-14. Picture of Lateral Acceleromete r, Longitudinal Accelerometer, and Angular Rate Sensor Installed Undern eath the Front Passenger Seat. When the carpet is replaced and the passenge r seat is in its normal position, the sensors are completely hidden.

PAGE 57

45 The brake and turn signals were meas ured by wiring the positive and negative (ground) leads of each in parallel with the inputs to their respective modules. All the signals measure approximately 12 volts when in use. A logic statement was implemented in LabVIEW that determines if the signal is above 6 volts. The result is a 1 corresponding to the signal being activated and a 0 for it not being activated. The event trigger was wired in the same was as the other signals ex cept the positive 12 volts had to be supplied. The event trigger is simply a pushbutton connect ed to the carÂ’s posit ive 12 volts and an input to a module. The other input to the modul e is connected to the carÂ’s frame (ground). The same logic used for the other signal s was also used for the event trigger. The GPS antenna was attached to the top of the trunk using the magnetic mount. It was wired to a serial/USB converter wh ich was connected to the USB port of the laptop. The positive lead of the power inverter was connected to the positive terminal of the carÂ’s battery and the nega tive lead was connected to th e vehicle chassis (ground). A 40-ampere fuse was installed on the positive lead near the battery to protect the inverter and all of the equipment connected to it. The multiplexer and mini-DV recorder was located in the trunk of the vehicle with the rest of the equipment. The four cam eras were installed in the ceiling of the vehicle cab. They came with brackets which al lowed them to be installed by screwing the bracket into the rigid foam ceiling. The or ientation of the cameras is completely adjustable. Each one was wired with 12 vo lt DC power and a composite video cable. A picture of the camera which monitors the driv erÂ’s eye movements is provided in Figure 415.

PAGE 58

46 Figure 4-15. Picture of Miniature Vi deo Camera and Installation Bracket. The video cable from each camera was wired to an input of the multiplexer. The output of the multiplexer was wired to the input of the mini-DV recorder using a composite video cable. All of the wires for the various sens ors, cameras, and power inverter were concealed beneath the door sills, carpet, and headliner. The only evidence inside the vehicle cab that would indicate the presence of special equipment is the four video cameras. All of the cameras are mounted to th e ceiling and although they are visible they should not interfere with the driverÂ’s field of view. A wiring diagram of the entire system is presented in Figure 4-16. The figure show s how the various components are connected and the cables that were used. Figure 4-17 is also provided to show the location of the components in the vehicle.

PAGE 59

47 Figure 4-16. Wiring Diagram of All System Components.

PAGE 60

48 Figure 4-17. Wiring Diagram Indicating the Lo cation of All System Components in the Vehicle.

PAGE 61

49 4.4 Calibration The entire system except for the GPS recei ver, event trigger, and brake and turn signals was setup in a laborat ory environment. The LabVIEW program used to operate the system was designed and refined in the laboratory. The sensitivity for the accelerometers provided by the technical sp ecification sheets was verified using the earthÂ’s gravity. Since they are static acceler ometers, they will measure exactly 1 g when placed on a level surface. There was no direct way of calibrating the angular rate sensor. However, it was verified that the sensor m easured 0 radians per second when stationary and provided positive and negative voltage for clockwise and counter clockwise rotations respectively. The laser tachometer and string potentiometer needed to be calibrated after being installed in the vehicle. The laser tachometer was calibrated by driving at a constant speed and comparing the result with the speed acquired by GPS. It was then possible to calculate the factor which relate s rotational frequency of the drive shaft in hertz to vehicle speed in miles per hour. The string potentiometer was calibrated by measuring the voltage with the wheel centered, turned one rotation to the left, and turned one rotation to the right. There was no calibration needed for the event trigger, brake and turn signals, or the GPS receiver. 4.5 Operation A set of instructions was created such that the user could oper ate the system after receiving minimal training. The first step is to turn on all of the equipment including the power inverter, laptop, power supplies, and vi deo equipment. Next, the user opens the LabVIEW program and GPSLog. The data acquis ition system is activated by pressing a start button on the control panel of the La bVIEW program. GPSLog connects to the GPS receiver and then starts logging the data. Th e video recording starts recording by pressing

PAGE 62

50 record on the min-DV recorder. Once the sy stem is running, the dr iving evaluation can begin. At the end of the evaluation the program s are terminated and the video recording is stopped. Finally, all of the equipment is turned off. The data from a one hour and twenty minute evaluation is approximately 5 megabyt es and is easily tran sferred to another computer using a USB flash drive.

PAGE 63

51 CHAPTER 5 SYSTEM VERIFICATTON Once the system was installed in the vehicle, it was necessary to make sure that all the components were working correctly and th at the data provided was accurate. Before implementing the system in actual research, a series of tests were performed to quantify the capabilities of the system and build a m odel for analyzing the data. It was decided that the data should be presented in a visual format so that the trends and interactions between measurements could be easily obs erved. The next section discusses the interpretation of the data and is followed by an analysis of the tests used to verify the systemÂ’s performance. 5.1 Interpretation of Data As previously mentioned, the data needs to be represented in a visual format in order to analyze the performa nce of the system. The independent variable for all the measurements is time and it is plotted along the horizontal axis. The measurements are made using a mixture of units so a conversion is necessary that will allow the full range of each measurement to be graphed on a single plot. The values for acceleration (measured in g Â’s) and yaw (measures in radians/sec ond) were observed to be less than a magnitude of one for most driving situations Since there are three measurements whose values are in the same range in their native units, the rest of the measurements are normalized to fit in this range. The steering wh eel position can be represented in radians, degrees, or revolutions. The d ecision was made to measure steering wheel position in revolutions because it is on the same order as the accelerations and yaw. Furthermore,

PAGE 64

52 measuring it in revolutions allows an invest igator to have a direct physical understanding of the wheel position. Whether sp eed was measured in kilome ters per hour or miles per hour, it was too large to fit in the zero to one range. Since th e FHWA research is focused on the time before, during, and after an inte rsection, vehicle speeds will most likely be less than 40 miles per hour. Ther efore the units for speed ar e represented in miles per hour divided by 20 (MPH/20). The brake and turn signals are simply on or off so they can easily be represented in the zero to one range. In order to simplify the plots, the brake and turn signal are only plotted when they are active. In addition to being represented in differe nt units, the different measurements are also measuring quantities in different directions (e.g., longitudinal and lateral acceleration). Therefore a sign convention must be develope d so that the different measurements can be plotted on a single graph and interprete d correctly. The sign convention is based on the assumption that th e vehicle is always moving in the forward direction. Longitudinal acceler ation will be positive for acceleration and negative for deceleration. Speed will always be measured in a positive sense since it always had a positive value. Lateral acceleration, yaw, a nd steering wheel position will be positive when the vehicle is moving forward and maki ng a right turn and negative when it is moving forward and making a left turn. The br ake status is positive when active and is assigned a value of one simply to keep it out of the way of the other measurements on the plot (i.e., non-zero). The turn signals will have a magnitude of 0.4 when active and will be positive for right turns and negative for left turns. Basically, the measurements related to forward movement (i.e., longitudinal accel eration and speed) will be the same sign regardless of the direction of a turn. The turn-related measurements (i.e., lateral

PAGE 65

53 acceleration, yaw, and steering wheel positi on) will be positive for right turns and negative for left turns. The si gn convention used when the vehi cle is in reverse will not be discussed here because there are not any reversing maneuvers in the FHWA research study. The GPS and vehicle dynamics data are record ed using two different programs that start recording at different times. Therefore, the two sets of data must be synchronized. The xcorr (cross-correlation) functi on in MATLAB is used to determine how much the GPS data needs to be shifted in order to synchronize it with the vehicle dynamics data. The laser tachometer has a lower threshold of about 4 miles per hour and cannot measure any speeds lower than this value. Since the speed provided by the laser tachometer and GPS are both sampled once per second, the GPS speed is used for all plots and calculations since it provides the necessary data in the 0 to 4 miles pe r hour range. An example of a maneuver plot for a left turn is provided in Figure 5-1. The abbreviations used for the measurements and their units as depicted in the figure are li sted in Table 5-1. Table 5-1. Measurement Abbreviations and Respective Units for All Plots. Measurement AbbreviationUnits Lateral Acceleration LA g Longitudinal (Forward) Acceleration FA g Steering Wheel Position ST revolutions (off of center) Yaw YW radians/second Speed (Laser Tachometer) SPT MPH / 20 Speed (GPS) SPG MPH / 20 Turn Signal Status TS ON/OFF Brake Status BR ON/OFF

PAGE 66

54 Figure 5-1. Example Plot of a Typical Left Turn. Figure 5-1 shows all of the measurements repr esented on a single plot. In order to better understand and interpret th e values of the result s, several simple driving maneuvers were performed. All of the maneuvers were performed in an empty parking lot with lane markings. The four maneuvers were (1) accel erating from a stop, (2) decelerating to a stop, (3) 90-degree right turn, and a (4) 90-degree left turn. Each maneuver was performed in two distinct manners. The mane uver was performed once in a normal, safe manner. The maneuver was then performed a second time in an aggressive manner, almost pushing the vehicle to its limits. 5.2 Analysis of Acceleration Test Maneuver For the acceleration test maneuvers, the ve hicle was at rest and accelerated until the vehicle reached 25 miles per hour. For the nor mal driving condition case, the accelerator pedal was gently depressed and the car acceler ated gradually. For the aggressive driving

PAGE 67

55 condition, the accelerator pedal was abruptly a pplied and the car was accelerated quickly but not enough for the tires to lose tracti on. The plots for the normal and aggressive acceleration test maneuvers are provided in Fi gures 5-2 and 5-3 respectively. For the rest of the analyses in this chapter, all of the plots will use the speed measured with GPS and will not display turn signal data. Figure 5-2. Plot of a Typical Norm al Acceleration Test Maneuver. From the plot of the normal acceleration, it can be seen the brake is initially applied and then released when the car begins to accelerate. The lateral acceleration, yaw, and steering wheel position are negligible as e xpected during a straight-line maneuver. The longitudinal acceleration reaches a maximum value of 0.18 g and then reduces gradually. When comparing the two scenarios, the aggr essive case achieves the top speed much quicker and has a higher l ongitudinal acceleration of 0.40 g The time needed to reach 25

PAGE 68

56 MPH was roughly 8 seconds for the normal case and 4.5 seconds for the aggressive case. The results from these tests are as expected a nd help to quantify the carÂ’s performance. If a driver were to accelerate with an acceleration of 0.4 g then it can be assumed that the driver is driving aggressive ly and at the carÂ’s limits. Figure 5-3. Plot of a Typical Aggres sive Acceleration Test Maneuver. 5.3 Analysis of Deceleration Test Maneuver A similar analysis was performed for the deceleration test. The car was decelerated in a straight-line from about 25 MPH. The pl ots for the decelerati on test maneuvers are provided in Figures 5-4 and 5-5. The normal deceleration generated a maximum force of about -0.19 g while the aggressive decel eration reached nearly -0.43 g The normal case took approximately 6 seconds to decelerate to a stop from 20 MPH. On the other hand, the aggressive case required only 3.5 seconds. Again, these results are as expected and assist in quantifying the deceler ation capabilities of the car.

PAGE 69

57 Figure 5-4. Plot of a Typical Norm al Deceleration Test Maneuver. Figure 5-5. Plot of a Typical Aggres sive Deceleration Test Maneuver.

PAGE 70

58 5.4 Analysis of Right Turn Test Maneuver The right and left turn test maneuvers were performed in a similar manner. The same roadway lane markings were used to ensu re the geometry of the turns is identical. In all cases, the driver approached the tur n, applied the brake, and then completed the turn. The only difference between the normal and aggressive scenarios was the speed at which the turn was taken. The speed during the midpoint of the turn was about 10 MPH for the normal case and 15 MPH for the aggressive case. The turns are tight and resemble those of a residential neighborhood. So, taking one of the turns at 15 MPH was actually close to the carÂ’s limits. The plots for the norma l and aggressive right turn test maneuvers are provided in Figures 5-6 and 5-7 respectively. Figure 5-6. Plot of a Typical Norm al Right Turn Test Maneuver.

PAGE 71

59 Figure 5-7. Plot of a Typical Aggres sive Right Turn Test Maneuver. In both cases, the longitudinal acceleration and speed relationship is as expected based on the acceleration and deceleration test cases. The speed decreases when the longitudinal acceleration is negativ e and the brake is applied. Th e speed begins to rise at the end of the turn when th e longitudinal acceleration is positive. The most relevant measurements for the turning maneuvers are lateral acceleration, ya w, and steering wheel position. The profiles for the th ree measurements are simila r however the magnitudes of lateral acceleration and yaw di ffer significantly. The steering wheel position reaches a maximum value of 0.695 revolutions for th e normal case and 0.825 revolution for the aggressive case. These two values should be approximately the same because the geometry of the turn is the same for both cas es and consequently the path traveled by the vehicle should also be the same. The reas on for the difference in the steering wheel position values can be determined by examining the profiles. For the normal case, the

PAGE 72

60 profile looks nearly symmetric and is with in 10 percent of the maximum value for 1.6 seconds. For the aggressive case, the steeri ng wheel position reaches its maximum value more abruptly and is within 10 percent of this value for only 0.9 seconds. Therefore, it can be inferred that for the normal case the dr iver turned the wheel gradually as they entered the turn, then held the wheel in a fixed position during the turn, and finally gradually turned the wheel back to center at the exit of the turn. On the other hand, for the aggressive case, the driver turned the wheel abruptly to begin the turn and almost immediately began returning it to center. Sin ce the aggressive turn was completed at a higher speed, it is expected that there will be higher values for yaw and lateral acceleration. This can be verified by examini ng the plots. The latera l acceleration reaches a maximum value of 0.332 g for the normal case and 0.561 g for the aggressive case. Yaw reaches a maximum value of 0.531 radians/second for the normal case and 0.685 radians/second for the aggressive case. A similar discussion will now be presented for the left turn test maneuvers. 5.5 Analysis of Left Turn Test Maneuver The results from the left turn test maneuve rs are nearly identical to those from the right turn test maneuvers. The plots for th e normal and aggressive left turn test maneuvers are provided in Figures 5-8 and 59 respectively. The same trends that were observed for the right turns test maneuvers are also observed for the left turns. The steering wheel position reaches similar maxi mum values of 0.71 and 0.77 revolutions for the normal and aggressive cases respectivel y. The maximum lateral acceleration achieved 0.248 g for the normal case and 0.603 g for the aggressive case. Yaw reached a maximum value of 0.423 radians/second for the nor mal case and 0.682 radians/second for the aggressive case.

PAGE 73

61 Figure 5-8. Plot of a Typical Norm al Left Turn Test Maneuver. Figure 5-9. Plot of a Typical Aggres sive Left Turn Test Maneuver.

PAGE 74

62 These results are as expected and confir m the results from the right turn test maneuvers. It can be assumed that if a driv er achieves a lateral acceleration value near 0.6 g for any driving maneuver, they are driving the car at its limits. However, a similar statement cannot be made for the yaw value. This is because yaw depends on the curvature of the turn (i.e., geometry of the intersection). Furthermore, the results for right and left turns measured on a road course will be different because in general right turns will have a smaller radius than left turns. This analysis provides a basis for understa nding the results obtained from the road course evaluations. Most impor tantly, the capabilities of the test vehicle have been quantified in terms of longit udinal and lateral acc eleration. If a driv er reaches these limits, then they are most likely driving in an unsafe manner. Although this was a general analysis, a more in-depth analysis could be performed for any particular driving maneuver depending on the hypothesis being teste d. For example, if a research study was focused on driving behaviors during lane cha nges, then several lane changes could be performed at different rates and vehicles speeds for a better understanding of the maneuver.

PAGE 75

63 CHAPTER 6 FHWA RESEARCH ANALYSIS This chapter discusses the procedures us ed to analyze the data for the FHWA research study. The purpose of the FHWA res earch is to study the effectiveness of intersection improvements recommended by th e FHWA. Participants navigate a road course and their driving performance is evaluated for ten intersections. The ten intersections are five pairs of similar inte rsections where one inte rsection features the recommended improvement and the other on e does not. The naming convention for the intersection uses the format 1A, 1B, 2A, 2B, etc. where the number refers to the intersection pair and A and B represent improved and unimproved intersections respectively. The analysis of the road cour se data is conducted in three separate steps and there is a MATLAB M-file specific to each step. The MATLAB files are named main stats and sort respectively. The main file is used to locate the te n intersections using the GPS data. It also converts the voltage measurements from each sensor to their desired units and then creates an output file for each maneuver (intersection). The stats file is used to separate each maneuver into three phases (approach, turn, a nd recovery). It calculates statistics for each measurement of each phase of each ma neuver. The statistics are compiled and output files are generated for each of the three phases. Lastly, the sort file takes all of the statistic files from each subject and sorts th e data by maneuver rather than subject. The final result is fifteen files which contain the statistics for each phase of each intersection pair. Finally, SPSS for Windows statistics softwa re is used to analyze the results and

PAGE 76

64 compare the improved and unimproved intersec tions. All of the MATLAB codes for this chapter may be found in Appendix D. An almost identical analysis is perfor med on the simulator data. There are ten simulator scenarios that are replicas of the ten road course maneuvers. The simulator creates output files that are similar to the output files generated by the main program used to analyze the road course data. Similar versions of the stats and sort programs are used to analyze and format the data so that is in a form identical to that of the road course data. The exact same procedure is used to analy ze the statistics from the road course and simulator using SPSS. A more detailed disc ussion of each MATLAB file and the SPSS analysis will now be presented. 6.1 Description of main.m (Road Course) Before the file can be run, the data must first be entered into the MATLAB workspace. The vehicle dynamics data and the GPS data are both stored in commadelimited text files. The GPS data contains a header and text stings (timestamp from GPSLog) so MATLABÂ’s Import Data Wizard is used to add the data to the workspace. The Import Data Wizard recognizes text and auto matically sorts out the numerical data to be imported. The vehicle dynamics data is imported using the dlmread command. The first step in the program is to plot the route traveled by the vehicle using the GPS coordinates. A simple algorithm is implemente d to determine when the vehicle is at the start or end location for a part icular intersection. The desired start and end locations were chosen by the driving evaluato rs. Their GPS coordinates we re found by driving the road course and marking each location by using the event trigger. A plot of the entire road course from GPS coordinate s is shown in Figure 6-1.

PAGE 77

65 Figure 6-1. FHWA Road Course Generated Usin g GPS Coordinates with Intersections of Interest Highlighted and Labeled. The ten intersections for which drivers are ev aluated are highlighted and labeled at the start of each maneuver. This pl ot is generated to ensure th at the proper intersections are selected for further analysis. The vehicle dynamics data and GPS da ta are recorded using two different programs and consequently the two sets of data are not synchronized. A correlation analysis between the speed provided by GPS and the speed provided by the laser tachometer is performed. The correlation is us ed to find the time of fset between the two sets of data and the GPS data is adjust ed accordingly. A plot of each maneuver is generated to verify that the GPS data was ad justed correctly. An example of one of these plots is provided in Figure 6-2.

PAGE 78

66 Figure 6-2. Example of Maneuver Plot Used to Verify That The Ve hicle Dynamics Data and GPS Data Are Synchronized. If for some reason the two sets of data ar e not synchronized exactly, the user has the option to shift the GPS data by the desired number of seconds. The plots are regenerated and the process is repeated unt il the data is synchronized. The final step is to create output file s for each maneuver. The output files contain all of the measurements in their appropri ate units for the enti re duration of each maneuver. The files are created using the dlmwrite command in MATLAB. The result is a comma-delimited file for each maneuver which can easily be imported into MATLAB for further analysis.

PAGE 79

67 6.2 Description of stats.m (Road Course) The stats MATLAB file is responsible for determining the three phases (approach, turn, and recovery) of each maneuve r and calculating statistics for each phase. The first step is to import the data by reading the file s generated by the main file using the dlmread command. The next step is to determin e the three phases of each maneuver. This is done by implementing an algorithm wh ich uses the heading from GPS and vehicle yaw. The algorithm uses the heading informati on to determine when the vehicle is in the middle of the turn. It then analyzes the yaw data to find out when the yaw is less than 0.05 radians per second. Since the yaw is near its maximum during the turn, the last point before the turn when the yaw is still less than 0.05 radians per second is considered to be the start of the turn. The first point after th e turn when the yaw returns to less than 0.05 radians per second is considered to be the e nd of the turn. The point immediately before the start of the turn is cons idered to be the end of the approach phase and the point immediately after the end of the turn is considered to be th e start of the recovery phase. The start of the approach phase and the end of the recovery phase are determined by the start and end locations of the maneuver re spectively. Although the distances for the approach and recovery phases vary from maneuv er to maneuver, they will always be the same for the same maneuver because the G PS locations for the start and end of the maneuver are constant. For comparison with the simulator scenarios, the distances of the approach and recovery phases need to be equal to those of the simulator. This is accomplished by converting the latitude and longi tude to feet and then calculating the distance traveled during each phase. The appr oach and recovery phases can then be truncated accordingly to equal the distan ces traveled in the respective simulator scenarios.

PAGE 80

68 The stats program generates a plot for each maneuver so the user can verify that the turn phase was selected correctly. A large and bold cross is used to mark the start and end of the turning phase. An example of one of these plots is shown in Figure 6-3. Figure 6-3. Plot of Maneuver 4B Generated Using stats.m With Markers Indicating the Start and End of The Turning Phase. Once the selection of phases is verified, the program is used to calculate statistics for each phase of each maneuver. The twelve stat istics calculated for each measurement are minimum, maximum, mean, root-mean-squa re, integral, variance, median, range, skewness, kurtosis, sum/time, and sum/distance. The last two statistics listed are the sum of the respective measurement normalized by th e time and distance traveled respectively during the phase. These twelve statistic s are calculated for lateral acceleration, longitudinal acceleration, steering, yaw, sp eed, and combined acceleration. Combined

PAGE 81

69 acceleration is the absolute ma gnitude of the overall accelerat ion of the vehicle and is defined in Equation 6.1 2 2_ _ accel long accel lat accel com Eq. 6.1 The final result is 72 statistics for each phase Since there are three phases per maneuver, and a total of ten maneuvers, there are 2,160 statistics calculated for each subject. These statistics are compiled into comma-delimited text files using the dlmwrite command in MATLAB. There are three files, one for each phase, that co ntain the statisti cs for all ten maneuvers. In addition to these statistics, the user enters the subjectÂ’s age and gender. The age is used to determine what age group the subject is in depending if the driver is young or old (65 years and older). The age grou p, gender, and maneuver number is also included in the output files to as sist with the SPSS analysis. 6.3 Description of sort.m (Road Course) The purpose of the sorting program is to compile all of the statistic files and sort them by maneuver pairs. The program reads all of the statistic files from each subject and arranges them by maneuver number. The program generates fifteen comma-delimited text files. There is one file for each phase of each maneuver pair. In each file there is a row for each subject. The row contains subject number, age group, gender, and maneuver number. It is then followed by all of the statistics for the improved maneuver (i.e., A ) which is followed by all of the statistics for the unimproved maneuver (i.e., B ). The statistics are arranged in this order so that they can easily be analyzed using SPSS. 6.4 Description of sim.m and sim_sort.m The simulator software produces files near ly identical to those produced by the main.m program for the road course. A program titled sim.m is very similar to stats.m for

PAGE 82

70 the road course and calculates statistics for each phase of each maneuver. There are subtle differences between the two programs. For example, the simulator data is provided in different units compared to t hose used for the road course, so sim.m is used to convert the data to the appropriate units. Also, the al gorithm used to determine the phases of each maneuver is slightly different due to the f act that the heading for the simulator always goes from zero to plus (right turn) or minus (left turn) ninety degrees. This is different than the road course heading which is the actu al global heading of the vehicle and varies based on the orientation of the intersection. The program generates statistics files which are identical to those generated by the stats.m program for the road course. A sorting program named sim_sort.m is used to sort the simulator statistics files in much the same way as sort.m did for the road course st atistics. There are two minor changes that were made. First, the order of the maneuvers in the statistic files for the simulator is different from the order for the ro ad course, so the indexing used to sort them is changed appropriately. The other change is to account for an extra intersection named 1C This intersection is similar to 1B except that is has a different improvement feature recommended by the FHWA. This improvement could not be found in any intersections in Gainesville, FL and as a result could not be incorporated into the road course. The effect of the improvement wanted to be stud ied so a simulator scenario was created. The sim_sort.m program creates fifteen output files similar to those created for the road course. The only difference is that the files for the maneuver 1 intersections contain the data for three maneuvers instead of two. 6.5 Description of SPSS Analysis SPSS statistical analysis soft ware is used to analyze the results from the road course vehicle and simulator statistics. A ge neral lineal model (GLM) repeated measures

PAGE 83

71 analysis was performed. This provides analysis of variance when the same measurement is used several times on each subject or case. For this analysis, the within-subject variable is the improvement of the intersection. The between-subject variable is age and it is used to divide the population into two groups (young and old). This general linear model analysis can test the effects of both the between-subject factors and the within-subject factors as well as the interaction between the two factors. The specific options chosen for the analysis were descriptive statistics, estim ates of effect size, and observed power. Additionally, a plot of the estimated marginal means of a particular measure is generated. This plot has a line for each age group, a vert ical axis for the mean of the measurement, and a horizontal axis for the improvement stat us of the intersection. The plot provides a useful too for visually analyz ing the effects of each factor and the interaction between them. A sample output file for one phase of one maneuver and five statistics (maximum lateral acceleration, maximum longitudinal a cceleration, maximum combine acceleration, maximum yaw, and maximum speed) is provided in Appendix E. Clearly performing and analyzing the results of this analysis for all st atistics of all phases of all maneuvers would be extremely time consuming. The analysis calculates the significance of the effect of each factor on a statistic and is used to determine which sta tistics are different as a result of intersection improvement or age group. Once some of the significant statistics are found, they can be further analyzed to infer about the effect of the intersection improvement or age group. For example, if both age groups had a higher maximum speed for the improved intersection, it could be inferred that the improvement helped by allowing the majority of participants to achieve a higher speed during the turn. This entire

PAGE 84

72 procedure reduces the time need ed for analysis by analyzing a particular statistic for all subjects at the same time rather than on an individual basis.

PAGE 85

73 CHAPTER 7 RESULTS AND DISCUSSION 7.1 FHWA Research Results At the time of this thesis submission, the FHWA research study had recently concluded. It is currently in the initial stag es of data analysis. Th e turning phase was the first phase to be analyzed because it capture s the dynamics during the intersection which is the main focus of interest. The approach and recovery phases will be analyzed at a later date to determine driver stability before and after the intersections. There are 72 statistics calculated for each phase, but it would be too time consuming to analyze all of the statistics for each maneuver using SPSS. The fi rst statistics to be investigated for the turning phase of all maneuvers were (1) ma ximum combined acceleration, (2) maximum longitudinal acceleration, (3) maximum latera l acceleration, (4) maximum yaw, and (5) maximum speed. SPSS can be used to determ ine if these statistics are significantly different for older and younger drivers (age group), improved and unimproved intersections (intersection improvement), and the interaction between age group and intersection improvement. An example analysis of these five statistics for one of the FHWA road course maneuvers wi ll now be presented. 7.2 Example Analysis of Turning Phase of FHWA Maneuver The five statistics are compiled into ta bles for each maneuver. The average and standard deviation for each statistic is tabul ated for four categories. The four categories are improved-young, improved-old, unim proved-young, and unimproved-old, where improved/unimproved refers to the intersecti on and young/old refers to the driverÂ’s age

PAGE 86

74 group. The statistics for Maneuver 2 from th e FHWA road course are presented in Table 7-1. Table 7-1. Summary of Statistics for Turnin g Phase of FHWA Road Course Maneuver 2. Maximum Combined Acceleration (g) Maximum Longitudinal Acceleration (g) Maximum Lateral Acceleration (g) Maximum Yaw (radians/sec) Maximum Speed (mph) Descriptive statistics Intersection Age Ave SD Ave SD Ave SD Ave SD Ave SD Young 0.30 0.05 0.19 0.06 0.25 0.06 0.40 0.06 22.172.66 Improved Old 0.30 0.06 0.19 0.08 0.22 0.06 0.37 0.07 20.752.91 Young 0.33 0.05 0.19 0.04 0.31 0.05 0.46 0.05 19.484.47 Unimproved Old 0.33 0.05 0.18 0.04 0.32 0.06 0.46 0.05 18.446.30 Inferential Statistics F p F p F p F p F p Age Group (Older Versus Younger Drivers) 0.12 0.73 0.11 0.74 0.49 0.49 1.29 0.26 3.03 0.09 Intersection Type (Improved versus Unimproved) 19.5 0.01*0.21 0.65 71.500.01*59.33 0.01*11.530.01* Interaction (Age x Intersection) 0.51 0.47 0.17 0.68 4.45 0.04*2.04 0.16 0.07 0.80 Maneuver 2 of the FHWA road course is a 90-degree right turn ont o a multi-lane divided highway. Both the improved and unimproved intersections are signalized. The improved intersection features right-turn channeliza tion. The right-turn channelization is a dedicated lane for all vehicles that are making a right turn at the intersection. It extends all the way into the receiving highway and allo ws drivers to accelerate up to speed before merging into traffic. A picture of the right-tu rn channelization is pr ovided in Figure 7-1.

PAGE 87

75 Figure 7-1. Picture of Right-Turn Channe lization from FHWA Road Course Maneuver 2A. An analysis of the statistics for FHWA Road Course Maneuver 2 will now be presented. First, the significance, if any, of each variable is investigated. There are four measurement statistics that are significant (p 0.05) for the type of intersection (improved/unimproved). They are maximum combined acceleration, maximum lateral acceleration, maximum yaw, and maximum speed The improved intersection has lesser values for all of these statistics excep t maximum speed. This can be explained by examining the equations for lateral accelerati on and yaw which are provided in Equations 7-1 and 7-2 respectively. r v al 2 Eq. 7-1

PAGE 88

76 r v Eq. 7-2 where v is the vehicle velocity and r is the radius of curvature. It should be noted that these equations are derived from their governing differential equations and in their current form are valid only for an arc of constant curvature. However, they can still be used to unders tand the interaction between speed, lateral acceleration, and yaw. Since the speed is gr eater for the improved intersection and the yaw and lateral acceleration is less, it can be inferred that the radius of curvature is greater than for the unimproved intersection. Th is can be easily verified by examining the geometry of each intersection. The results indicate that driv ers are able to negotiate the improved intersection at higher speeds while experiencing less lateral forces. Maximum lateral acceleration is the only variable that is significant for the interaction between age group and intersection type. The older drivers had a greater value of maximum lateral acceleration for the unimproved intersection and a smaller value for the improved intersection. Si nce the maximum speed is great er for the younger age group for both intersections, we would expect the respective values for maximum yaw to also be greater. The only explanation is that th e maximum statistic measures do not indicate the time at which they occurred. If it is a ssumed that both age groups of drivers achieved maximum lateral acceleration at approximately the same time (midway through the turn). Then it is likely that the younger and older drivers reached their maximum speed at different times. Further analysis of the data could reveal if the older drivers began to decelerate earlier. Maximum speed is the only variable that approaches significance (p 0.10) for the age group. Younger drivers had greater speed for both the improved and unimproved

PAGE 89

77 intersections. This is an expected result si nce older drivers generally drive with more caution and at a lower speed. This analysis, or one very similar can be performed on the remaining maneuvers. Examination of all 72 measurement statistics can reveal which ones are most significant. The measurement statistics of interest may vary depending on the specific improvement being examined in each maneuver. Furthermor e, the statistics can be separated into different cases depending on the driving cond itions (i.e., stopping before the turn or signal status). The other phases (approach a nd recovery) of each maneuver will also be analyzed using a similar process. The approach and recovery phases will be analyzed to determine how each intersection affects the drive rÂ’s stability before and after the turn. It has been shown that the measur ement statistics and their associ ated significance is useful in analyzing each intersection. These result s combined with an understanding of the intersection can be used to infer inform ation about the driverÂ’s performances.

PAGE 90

78 CHAPTER 8 CONCLUSION The vehical dynamics acquisition system developed for the FHWA study was able to successfully measure and record the vehicl e dynamics of a Buick sedan. The addition of GPS allowed the system to acquire the vehicle dynamics at specific locations. The system is flexible and can measure a wide va riety of inputs. It can easily be expanded with the addition of SCC modules and transducers to measured additional information. The system was first used for the FHWA study and proved to be a useful tool in quantifying driving behavior. It could be used in its current form for a different study concerned with a different driving behavior The only change would be to the MATLAB code in order to analyze the particular area of interest. The video equipment, although not directly incorporated with the data acquisit ion and GPS equipment, was a helpful tool for the evaluators. It was used to train new eval uators and to review evaluations to ensure that evaluators were scoring driving errors consistently. 8.1 Future Research Studies It is likely that this system will be used in the future for other driving behavior studies. If needed, additional sensors can eas ily be added. For example, a load cell could be used to measure the amount of force the driver is applying to a pedal. The system could even be adapted to measure physiological responses from the driver such as heart rate, or perspiration. Also, the cameras can be remounted to change the field of view. For example, the cameras could be used to mon itor vehicle lane maintenance by aiming the cameras toward the lane markings next to the vehicle.

PAGE 91

79 8.2 Recommendations and Improvements Although the system worked well and as intended, some improvements could be made that would make it easier to use and operate. The first recommendation is to find an alternate way to measure the vehicle speed. The speed obtained from the GPS was only updated once per second and was less accurate when the vehicle was stationary or moving slowly. The speed measured using the laser tachometer was inconsistent due to the laser not being reflected sometimes. Furt hermore, the laser t achometer could not measure speed below 5 miles per hour. An infrared ground speed sensor like the Correvit LF™ ( Appendix B) would be able to measure vehicle speed faster and more accurately. The second recommendation is to use only one program for operating the system. This could be done by incorporating the GPS w ith the rest of the system via a serial port and LabVIEW. The LabVIEW Virtual Instrume nt could be programmed to interface with the GPS antenna using the NMEA 0183 Standard. The final recommendation is related to th e overall design and installation of the system. If a similar system were to be in stalled in another vehi cle then all of the equipment should be contained in a single co ntainer or rack. The r ack would contain the power inverter, all power supp lies, computer, and data acqui sition system. There should be one panel that contains all of the conn ections for the sensors. Also, a personal computer (PC) could be used instead of a laptop computer. This would allow the hardware to be contained in the rack and th e monitor located in th e cab of the vehicle. The computer could be controlled using a Bluetooth keyboard and/or mouse. The advantage of designing the system like this is that it would take up less room in the vehicle, be simpler to operate, and easier to install or remove.

PAGE 92

APPENDIX A INTERSECTION DIAGRAMS

PAGE 93

81

PAGE 94

82

PAGE 95

83

PAGE 96

84

PAGE 97

85

PAGE 98

86

PAGE 99

87

PAGE 100

88

PAGE 101

89

PAGE 102

90

PAGE 103

91

PAGE 104

92 APPENDIX B COMPONENT TECHNICAL SPECIFICATIONS Kistler Static Accelerometers Model Number 1. 8310A2 2. 8310A10 3. Range 4. 2 g 5. 10 g 6. Sensitivity 7. 1000 mV/g 8. 200 mV/g 9. Threshold 1 0. 540 g 1 1. 2830 g 1 2. Frequency Response 1 3. 0-300 Hz 1 4. 0-180 Hz 1 5. Temp. Operating Range 1 6. -40 – 85 C 1 7. -40 – 85 C 1 8. Supply Voltage 1 9. 3.8 – 16 volts 2 0. 3.8 – 16 volts 2 1. Supply Current (nom.) 2 2. 1.3 amps 2 3. 1.3 amps 2 4. Housing 2 5. Titanium, hermetically sealed 2 6. Titanium, hermetically sealed 2 7. Connector 2 8. 4-pin Microtech 2 9. 4-pin Microtech

PAGE 105

93 BEI GyroChip Angular Rate Sensor 3 0. Model Number 3 1. AQRS-00075-111 3 2. Range 3 3. 75/sec 3 4. Sensitivity 3 5. 33.33 /volt 3 6. Input Voltage 3 7. + 5 volts 5% 3 8. Input Current 3 9. <20 mA 4 0. Bias Calibration 4 1. +2.50 volts (nom.) 4 2. Temp. Operating Range 4 3. -40 – 85 C 4 4. Connector 4 5. 3-pin LaserTach Laser Tachometer Speed Range 30,000 RPM max. Input Voltage 18 – 30 volts Input Current 3 – 20 mA (constant) Output Type Square wave Output Amplitude 1 Vpp @ 4 mA supply current Operating Range 6 feet (90) 2 feet (45) Temp. Operating Range -40 – 85 C Connector BNC

PAGE 106

94 Celesco Cable-Extension Positi on Transducer (String-Pot) (* indicates custom selection) 4 6. Model Number 4 7. PT1DC-50-BK-Z10-MC4-SG 4 8. Range* 4 9. O – 50 inches 5 0. Accuracy 5 1. 0.15% (full scale) 5 2. Resolution 5 3. infinite 5 4. Input Voltage 5 5. 14.5 – 40 volts 5 6. Input Current 5 7. 10 mA max. 5 8. Output Voltage* 5 9. 0 – 10 volts 6 0. Cable 6 1. 0.019-in. dia.nylon-coated stainless steel 6 2. Cable Exit* 6 3. back 6 4. Cable Tension 6 5. 5 oz. 6 6. Temp. Operating Range 6 7. -40 – 200 F 6 8. Connector* 6 9. 4-pin micro-connector (1 not used)

PAGE 107

95 APPENDIX C CUSTOM PULLEY DRAWINGS

PAGE 108

96

PAGE 109

97 APPENDIX D MATLAB CODES main.m % ********************************************************************** % % FHWA MAIN (ROAD COURSE) % % University of Florida % College of Publich Health and Health Professions % Occupational Therapy Department % Created by: Ethan J. Davis (Mechanical & Aerospace Engineering) % Date: 06-27-05 % % ------------------------INSTRUCTIONS -----------------------------% Place this M-File in the same folder with "data.txt" and "GPSLog.txt" % Open Matlab, Open this M-File % In main MATLAB menu, Click on "File" --> "Import Data..." % Locate the Folder and select "GPSLog.txt", press "Next" then "Finish" % Run this M-File (F5) % % ********************************************************************** format short g % sort GPS data data = GPSLog; lat = data(:,1); % latitude long = data(:,2); % longitude sp_gps = data(:,3)*1.15077945; % speed (GPS)/convert from knots to mph head = data(:,4); % heading dt = 0.1; % time step (0.1 sec = 10 Hz) A = ['2A6B1A4B4A6A1B2B3B3A']; % text string of maneuver labels % start/end GPS locations and headings for each maneuver locs = [29.70983498891190, -82.35401500066121, 13; % Start 2a 29.70910663604740, -82.35062166849770, 134; % End 2a 29.70412502288820, -82.36198666890461, 244; % S 6b 29.7033, -82.36, 90; % E 6b 29.68863665262860, -82.36857500076290, 271; % S 1a 29.683, -82.3725, 180; % E 1a 29.65938498179120, -82.34601332346600, 271; % S 4b 29.6575, -82.34730332692460, 182; % E 4b 29.65943164825440, -82.35261001586910, 270; % S 4a 29.6578, -82.35378999710080, 180; % E 4a 29.65947335561120, -82.36929833094280, 270; % S 6a 29.656, -82.37249832153320, 180; % E 6a 29.67413164774580, -82.39865334828690, 270; % S 1b 29.670, -82.4015, 180; % E 1b 29.662, -82.40154666900639, 180; % S 2b 29.6596, -82.405, 272; % E 2b 29.65189666748050, -82.37493000030520, 88; % S 3b 29.65512332916260, -82.37237332661950, 360; % E 3b 29.6886, -82.3779, 90; % S 3a 29.693, -82.3723, 1]; % E 3a lo_tol = 0.0005 ; % low tolerance for finding maneuver locations hi_tol = 0.0005; % high tolerance " " hd_tol = 4; % heading tolerance " " f = 0;clear('f');f = 0;x = 1;y = 1;null = 0; % for each start or end location... % find when vehicle is within prescribed range of location % and properly oriented for r = 1:length(locs) clear('f2') f2 = find(lat < locs(r,1)+hi_tol & lat > locs(r,1)-lo_tol & long < locs(r,2)+lo_tol & long > locs(r,2)-hi_tol & head < locs(r,3)+hd_tol & head > locs(r,3)-hd_tol); if size(f2) < 1 null(x) = r; % list of locations not found

PAGE 110

98 x = x+1; else f3 = find(f2 > max(f)); f(r) = min(f2(f3(1)):f2(f3(length(f3)))); c(y) = r; % list of locations found y = y+1; end end % the following steps are in the event that the computer crashed and the date % files are incomplete. it ensures that each maneuver has a start and end % location, otherwise they are not included. % if first found location is not a start location, remove it from list if rem(c(1),2) ~= 0 w = c; else w = c(2:length(c)); end clear('c'); c = w; % if last location is not an end location, remove it from list if rem(c(length(c)),2) ~= 0 w = c(1:length(c)-1); else w = c; end % plot maneuvers and road course using GPS data figure(1);clf for h = w(1):2:w(length(w))-1 % plot each maneuver in list lo = (f(h)); hi = (f(h+1)); plot((long(lo:hi)),(lat(lo:hi)),'Color',[1 0 0],'LineWidth',3); text_name = [A(h:h+1)]; if h == 1 long_factor = -0.003; else long_factor = 0; end % add maneuver labels to plot text((long(lo))+long_factor,(lat(lo))+.0015,text_name);hold on end plot(long,lat) % plot entire road course xlim([-82.41 -82.34]) ylim([29.648 29.715]) ylabel('Latitude [Decimal Degrees]') xlabel('Longitude [Decimal Degrees]') % Prompt user if maneuvers were selected correctly In_1 = input('Are the selected Maneuvers correct ? (1 = yes, 2 = no) '); if In_1 == 1 adjust = 0; adjust_index = 0; adjust_index_2 = 0; secs = 0; secs_index = 0; % open and read kinematic data file data2 = dlmread('data.txt',','); % sort kinematic data al = data2(:,4); % acceleration, lateral af = data2(:,5); % acceleration, forward sp = data2(:,1); % speed sw = data2(:,7); % steering wheel br = data2(:,10); % brake ar = data2(:,6); % angular rate lt = data2(:,2); % left turn signal rt = data2(:,3); % right turn signal trig = data2(:,9); % event trigger % remove DC offsets al = al mean(al); af = af mean(af); ar = ar mean(ar); % conversions al = al/1.007; % convert to g af = -af/0.198; % convert to g sw = -(sw+2.18)/2.92; % convert to revolutions ar = ar*.581776; % converr to rad/s % speed conversion for s = 1:length(sp) if sp(s) < 0.77 sp(s) = 0;

PAGE 111

99 elseif sp(s) > 14 sp(s) = 0; else sp(s) = sp(s)*4.5575424; % convert to mph end end % create new vector with every tenth sample of speed for i = 1:fix(length(sp)/10) sp2(i) = sp(i*10); end % boolean expressions for determining brake and turn signal status for k = 1:length(al) if br(k) == 1 & rt(k) == 1 & lt(k) == 0 lt(k) = 1;rt(k) = 0; elseif br(k) == 1 & rt(k) == 0 & lt(k) == 1 lt(k) = 0;rt(k) = 1; elseif br(k) == 1 & rt(k) == 1 & lt(k) == 1 lt(k) = 0;rt(k) = 0; end end t = dt:dt:length(br); % create time array [10 samp/sec] t2 = 0:1:length(br); % create time array for speed [1 samp/sec] lti = 0;rti = 0;bri = 0; % initialize left turn, right turn, and brake indices % correlate GPS speed and tachometer speed [y,lags] = xcorr(sp_gps(1:length(sp_gps)),sp2(1:length(sp2))); [c,I] = max(y); nlags = length(lags); nzero = nlags (nlags-1)/2; tau = I nzero; f = f tau; % do until speeds are correlated while adjust_index < 1 tau = tau + adjust; if tau < 0 sp_gps(1:-tau) = 0; k2 = -tau + 1; else k2 = 1; end temp_sp_gps = sp_gps; temp_lat = lat; temp_long = long; temp_head = head; for k = k2:length(sp_gps)-tau sp_gps(k) = temp_sp_gps(k+tau); lat(k) = temp_lat(k+tau); long(k) = temp_long(k+tau); head(k) = temp_head(k+tau); end tau = 0; for q = w(1):2:w(length(w)) % one maneuver at a time rangelo = (f(q))/dt; % start of maneuver rangehi = (f(q+1)+secs)/dt; % end of maneuver v = 1;u = 1;y = 1; % create indices for plotting turn signal and brake for z = rangelo:rangehi if rt(z) == 1 rt2(v) = 0.4;rti(v) = z*dt;v = v + 1; end if lt(z) == 1 lt2(u) = -0.4;lti(u) = z*dt;u = u + 1; end if br(z) == 1 br2(y) = 1;bri(y) = z*dt;y = y + 1; end end % plot the data for for each maneuver seperately from start location to end location figure(((q+1)/2)+1);clf plot(t(rangelo:rangehi),al(rangelo:rangehi),'Color',[0 0 0],'LineStyle','');hold on plot(t(rangelo:rangehi),af(rangelo:rangehi),'Color',[0 0 0],'LineStyle',':'); plot(t(rangelo:rangehi),sw(rangelo:rangehi),'Color',[0.5 1 0],'LineStyle',':'); plot(t(rangelo:rangehi),ar(rangelo:rangehi),'Color',[0.5 1 0],'LineStyle',''); plot(t2(rangelo/10:rangehi/10),sp2(rangelo/10:rangehi/10)/20,'Color',[0 0.5 1],'LineStyle','-.','LineWidth',2);

PAGE 112

100 plot(t2(rangelo/10:rangehi/10),sp_gps(rangelo/10:rangehi/10)/20,'Color',[0 0 1],'LineStyle','-','LineWidth',2); ylim([-1 1.5]) if lti ~= 0 plot(lti,lt2,'LineStyle','none','Marker','+','MarkerSize',3,'Color',[1,0,1]); else plot(rangelo*dt,0,'LineStyle','none','Marker','+','MarkerSize',3,'Color',[1,0,1]) end if rti ~= 0 plot(rti,rt2,'LineStyle','none','Marker','+','MarkerSize',3,'Color',[1,0,1]); else plot(rangelo*dt,0,'LineStyle','none','Marker','+','MarkerSize',3,'Color',[1,0,1]) end if bri ~= 0 plot(bri,br2,'LineStyle','none','Marker','x','MarkerSize',6,'Color','black'); else plot(rangelo*dt,0,'LineStyle','none','Marker','x','MarkerSize',6,'Color','black') end hold off legend('LA','FA','ST','YW','SPT','SPG','TS','TS','BR',0); xlabel('time [sec]') ylabel('[g], [rev], [rad/s], [mph/20]') Maneuver = ['Maneuver ']; TitleName = [Maneuver, A(q:q+1)]; title(TitleName) clear('rt2','rti','lt2','lti','br2','bri') rt2 = 0;rti = 0;lt2 = 0;lti = 0;br2 = 0;bri = 0; end % ask user if GPS speed and tachometer speed are correlated if adjust_index == 1 else In_2 = input('Are Tach Speed (purple) and GPS Speed (yellow) less than 1 second apart ? (1 = yes, 2 = no) '); In_2 = 1; end % if not, which way does GPS speed need to shift if In_2 == 1 adjust_index = 1; adjust = 0; else In_3 = input('Is the Tach speed (purple) to the right of the GPS speed (yellow) ? (1 = yes, 2 = no) '); if In_3 == 1 sign = -1; else sign = 1; end adjust = input('What is the difference in seconds ? '); adjust = adjust*sign; end % ask user if they would like to create output files In_5 = input('Would you like to create Output Files ? (1 = yes, 2 = no) '); if In_5 == 1 status = 'Please Wait...' % can take a few minutes for q = w(1):2:w(length(w)) lo = (f(q))/dt; hi = (f(q+1)+secs)/dt; % create new array with GPS data every 1/10 second for i = 10:10:length(lat)*10 lat3(i-9:i) = lat(i/10); sp_gps3(i-9:i) = sp_gps(i/10); long3(i-9:i) = long(i/10); head3(i-9:i) = head(i/10); end format long out = [t(lo:hi)',af(lo:hi),al(lo:hi),sw(lo:hi),ar(lo:hi),br(lo:hi),lt(lo:hi),rt(lo:hi),(sp_gps3 (lo:hi))',lat3(lo:hi)',long3(lo:hi)',head3(lo:hi)']; TitleName = [Maneuver, A(q:q+1)]; dlmwrite(TitleName,out) % create output file end end end % end while statement end % end if statement status = 'Finished'

PAGE 113

101 stats.m % ********************************************************************** % % FHWA Statisitics (ROAD COURSE) % % University of Florida % College of Publich Health and Health Professions % Occupational Therapy Department % Created by: Ethan J. Davis (Mechanical & Aerospace Engineering) % Date: 08-13-05 % % --------------------------INSTRUCTIONS -------------------------% Place this M-File in the same folder with the 10 Maneuver Files % Open Matlab, Open this M-File % Run this M-File (F5) % % ********************************************************************** clear all format short g heading = [33 224 241 241 240 240 200 68 335 70]; % heading during the middle of each turn A = ['2A6B1A4B6A1B2B3B4C3A']; % text string of maneuver labels Maneuver = ['Maneuver ']; % maneuver string app = 0; turn = 0; rec = 0; ttm = 0; sub = input('What is the Subject Number ? '); age = input('How old is the subject? '); gen = input('What gender is the subject? (0 = MALE, 1 = FEMALE) '); % determine age groups if age < 65 age_1 = 0; age_2 = 0; elseif age > 64 & age < 76 age_1 = 1; age_2 = 1; elseif age > 75 & age < 86 age_1 = 2; age_2 = 1; else age_1 = 3; age_2 = 1; end % process one maneuver at a time for k = 1:2:19 rec_flag = 0; TitleName = [Maneuver, A(k:k+1)]; clear('f','f2','f3','f4','f5','data','t','af','al','sw','ar','br','lt','rt','sp','lat','l ong','lat_ft','lonf_ft','head','t2','sp2','head2') data = dlmread(TitleName,','); if A(k+1) == 'A' imp = 1; elseif A(k+1) == 'B' imp = 2; elseif A(k+1) == 'C' imp = 3; else status = 'ERROR' end t = data(:,1); % time af = data(:,2); % forward accel al = data(:,3); % lateral accel sw = data(:,4); % steering ar = data(:,5); % yaw br = data(:,6); % brake lt = data(:,7); % left turn rt = data(:,8); % right turn sp = data(:,9)*10; % speed lat = data(:,10); % latitude long = data(:,11); % longitude head = data(:,12); % heading lat_ft = lat*364500; % convert to feet long_ft = long*316800; % convert to feet

PAGE 114

102 % Create new arrays for time and speed by taking every 10th sample. % This is to remove the stair-step appearance on the plots. for i = 1:length(t)/10 t2(i) = t(i*10); sp2(i) = sp(i*10); head2(i) = head(i*10); end % find when vehicle is in middle of turn if k == 1 | k == 13 % i.e. right turns f = find(head > heading((k+1)/2) 5 & head < heading((k+1)/2) + 50); else % i.e. left turns f = find(head > heading((k+1)/2) 50 & head < heading((k+1)/2) + 0); end f2 = find(abs(ar) < 0.05); f3 = find(f2 < min(f)); tsi = f2(max(f3)); % turn start index if k == 1 | k == 13 if k == 1 f6 = find(head > 100); else f6 = find(head > 230); end f7 = find(f6 > tsi); f4 = find(f2 > f6(min(f7))); else f4 = find(f2 > min(f)); end tei = f2(min(f4)); % turn end index rsi = tei + 1; % recovery start index rei = length(ar); % recovery end index asi = 1; % approach start index aei = tsi 1; % approach end index f5 = find(sp == 0); % determine if driver stopped if length(f5) > 0 stop = 1; else stop = 0; end % calculate distance traveled during approach phase app_dis = 0; for i = asi:aei-1 app_dis = app_dis + sqrt((lat_ft(i)-lat_ft(i+1))^2+(long_ft(i)-long_ft(i+1))^2); end % calculate distance traveled during turning phase turn_dis = 0; for i = tsi:tei-1 turn_dis = turn_dis + sqrt((lat_ft(i)-lat_ft(i+1))^2+(long_ft(i)long_ft(i+1))^2); end % calculate distance traveled during recovery phase (stop at 500 feet) rec_dis = 0; while rec_flag < 1 for i = rsi:rei-1 if rec_dis > 500 rei = i; rec_flag = 1; else rec_dis = rec_dis + sqrt((lat_ft(i)-lat_ft(i+1))^2+(long_ft(i)long_ft(i+1))^2); end end rec_flag = 1; end app_time = t(aei) t(asi); % time to complete approach phase turn_time = t(tei) t(tsi); % time to complete turning phase rec_time = t(rei) t(rsi); % time to complete recovery phase com = 0; for i = 1:length(al) com(i) = sqrt(af(i)^2+al(i)^2); % combined acceleration end % calculate statistics for each phase % approach if k == 1 | k == 13 % right turns factor = 1; app_yaw_min = min(ar(asi:aei));

PAGE 115

103 app_yaw_max = max(ar(asi:aei)); app_lat_min = min(al(asi:aei)); app_lat_max = max(al(asi:aei)); app_str_min = min(sw(asi:aei)); app_str_max = max(sw(asi:aei)); else factor = -1; app_yaw_min = -max(ar(asi:aei)); app_yaw_max = -min(ar(asi:aei)); app_lat_min = -max(al(asi:aei)); app_lat_max = -min(al(asi:aei)); app_str_min = -max(sw(asi:aei)); app_str_max = -min(sw(asi:aei)); end app_fwd_min = min(af(asi:aei)); app_fwd_max = max(af(asi:aei)); app_spd_min = min(sp(asi:aei)); app_spd_max = max(sp(asi:aei)); app_com_min = min(com(asi:aei)); app_com_max = max(com(asi:aei)); app_yaw_avg = mean(factor.*ar(asi:aei)); app_yaw_rms = norm(factor.*ar(asi:aei))/sqrt(length(ar(asi:aei))); app_yaw_int = sum(factor.*ar(asi:aei))*0.1*length(ar(asi:aei)); app_yaw_var = var(factor.*ar(asi:aei),1); app_yaw_med = median(factor.*ar(asi:aei)); app_yaw_ran = app_yaw_max app_yaw_min; app_yaw_skw = skewness(factor.*ar(asi:aei),0); app_yaw_kur = kurtosis(factor.*ar(asi:aei),0); app_yaw_s_t = sum(factor.*ar(asi:aei))/app_time; app_yaw_s_d = sum(factor.*ar(asi:aei))/app_dis; app_lat_avg = mean(factor.*al(asi:aei)); app_lat_rms = norm(factor.*al(asi:aei))/sqrt(length(al(asi:aei))); app_lat_int = sum(factor.*al(asi:aei))*0.1*length(al(asi:aei)); app_lat_var = var(factor.*al(asi:aei),1); app_lat_med = median(factor.*al(asi:aei)); app_lat_ran = app_lat_max app_lat_min; app_lat_skw = skewness(factor.*al(asi:aei),0); app_lat_kur = kurtosis(factor.*al(asi:aei),0); app_lat_s_t = sum(factor.*al(asi:aei))/app_time; app_lat_s_d = sum(factor.*al(asi:aei))/app_dis; app_fwd_avg = mean(af(asi:aei)); app_fwd_rms = norm(af(asi:aei))/sqrt(length(af(asi:aei))); app_fwd_int = sum(af(asi:aei))*0.1*length(af(asi:aei)); app_fwd_var = var(af(asi:aei),1); app_fwd_med = median(af(asi:aei)); app_fwd_ran = app_fwd_max app_fwd_min; app_fwd_skw = skewness(af(asi:aei),0); app_fwd_kur = kurtosis(af(asi:aei),0); app_fwd_s_t = sum(af(asi:aei))/app_time; app_fwd_s_d = sum(af(asi:aei))/app_dis; app_str_avg = mean(factor.*sw(asi:aei)); app_str_rms = norm(factor.*sw(asi:aei))/sqrt(length(sw(asi:aei))); app_str_int = sum(factor.*sw(asi:aei))*0.1*length(sw(asi:aei)); app_str_var = var(factor.*sw(asi:aei),1); app_str_med = median(factor.*sw(asi:aei)); app_str_ran = app_str_max app_str_min; app_str_skw = skewness(factor.*sw(asi:aei),0); app_str_kur = kurtosis(factor.*sw(asi:aei),0); app_str_s_t = sum(factor.*sw(asi:aei))/app_time; app_str_s_d = sum(factor.*sw(asi:aei))/app_dis; app_spd_avg = mean(sp(asi:aei)); app_spd_rms = norm(sp(asi:aei))/sqrt(length(sp(asi:aei))); app_spd_int = sum(sp(asi:aei))*0.1*length(sp(asi:aei)); app_spd_var = var(sp(asi:aei),1); app_spd_med = median(sp(asi:aei)); app_spd_ran = app_spd_max app_spd_min; app_spd_skw = skewness(sp(asi:aei),0); app_spd_kur = kurtosis(sp(asi:aei),0); app_spd_s_t = sum(sp(asi:aei))/app_time; app_spd_s_d = sum(sp(asi:aei))/app_dis; app_com_avg = mean(com(asi:aei)); app_com_rms = norm(com(asi:aei))/sqrt(length(com(asi:aei))); app_com_int = sum(com(asi:aei))*0.1*length(com(asi:aei)); app_com_var = var(com(asi:aei),1); app_com_med = median(com(asi:aei));

PAGE 116

104 app_com_ran = app_com_max app_com_min; app_com_skw = skewness(com(asi:aei),0); app_com_kur = kurtosis(com(asi:aei),0); app_com_s_t = sum(com(asi:aei))/app_time; app_com_s_d = sum(com(asi:aei))/app_dis; % turn if k == 1 | k == 13 factor = 1; turn_yaw_min = min(ar(tsi:tei)); turn_yaw_max = max(ar(tsi:tei)); turn_lat_min = min(al(tsi:tei)); turn_lat_max = max(al(tsi:tei)); turn_str_min = min(sw(tsi:tei)); turn_str_max = max(sw(tsi:tei)); else factor = -1; turn_yaw_min = -max(ar(tsi:tei)); turn_yaw_max = -min(ar(tsi:tei)); turn_lat_min = -max(al(tsi:tei)); turn_lat_max = -min(al(tsi:tei)); turn_str_min = -max(sw(tsi:tei)); turn_str_max = -min(sw(tsi:tei)); end turn_fwd_min = min(af(tsi:tei)); turn_fwd_max = max(af(tsi:tei)); turn_spd_min = min(sp(tsi:tei)); turn_spd_max = max(sp(tsi:tei)); turn_com_min = min(com(tsi:tei)); turn_com_max = max(com(tsi:tei)); turn_yaw_avg = mean(factor.*ar(tsi:tei)); turn_yaw_rms = norm(factor.*ar(tsi:tei))/sqrt(length(ar(tsi:tei))); turn_yaw_int = sum(factor.*ar(tsi:tei))*0.1*length(ar(tsi:tei)); turn_yaw_var = var(factor.*ar(tsi:tei),1); turn_yaw_med = median(factor.*ar(tsi:tei)); turn_yaw_ran = turn_yaw_max turn_yaw_min; turn_yaw_skw = skewness(factor.*ar(tsi:tei),0); turn_yaw_kur = kurtosis(factor.*ar(tsi:tei),0); turn_yaw_s_t = sum(factor.*ar(tsi:tei))/turn_time; turn_yaw_s_d = sum(factor.*ar(tsi:tei))/turn_dis; turn_lat_avg = mean(factor.*al(tsi:tei)); turn_lat_rms = norm(factor.*al(tsi:tei))/sqrt(length(al(tsi:tei))); turn_lat_int = sum(factor.*al(tsi:tei))*0.1*length(al(tsi:tei)); turn_lat_var = var(factor.*al(tsi:tei),1); turn_lat_med = median(factor.*al(tsi:tei)); turn_lat_ran = turn_lat_max turn_lat_min; turn_lat_skw = skewness(factor.*al(tsi:tei),0); turn_lat_kur = kurtosis(factor.*al(tsi:tei),0); turn_lat_s_t = sum(factor.*al(tsi:tei))/turn_time; turn_lat_s_d = sum(factor.*al(tsi:tei))/turn_dis; turn_fwd_avg = mean(af(tsi:tei)); turn_fwd_rms = norm(af(tsi:tei))/sqrt(length(af(tsi:tei))); turn_fwd_int = sum(af(tsi:tei))*0.1*length(af(tsi:tei)); turn_fwd_var = var(af(tsi:tei),1); turn_fwd_med = median(af(tsi:tei)); turn_fwd_ran = turn_fwd_max turn_fwd_min; turn_fwd_skw = skewness(af(tsi:tei),0); turn_fwd_kur = kurtosis(af(tsi:tei),0); turn_fwd_s_t = sum(af(tsi:tei))/turn_time; turn_fwd_s_d = sum(af(tsi:tei))/turn_dis; turn_str_avg = mean(factor.*sw(tsi:tei)); turn_str_rms = norm(factor.*sw(tsi:tei))/sqrt(length(sw(tsi:tei))); turn_str_int = sum(factor.*sw(tsi:tei))*0.1*length(sw(tsi:tei)); turn_str_var = var(factor.*sw(tsi:tei),1); turn_str_med = median(factor.*sw(tsi:tei)); turn_str_ran = turn_str_max turn_str_min; turn_str_skw = skewness(factor.*sw(tsi:tei),0); turn_str_kur = kurtosis(factor.*sw(tsi:tei),0); turn_str_s_t = sum(factor.*sw(tsi:tei))/turn_time; turn_str_s_d = sum(factor.*sw(tsi:tei))/turn_dis; turn_spd_avg = mean(sp(tsi:tei)); turn_spd_rms = norm(sp(tsi:tei))/sqrt(length(sp(tsi:tei))); turn_spd_int = sum(sp(tsi:tei))*0.1*length(sp(tsi:tei)); turn_spd_var = var(sp(tsi:tei),1); turn_spd_med = median(sp(tsi:tei)); turn_spd_ran = turn_spd_max turn_spd_min;

PAGE 117

105 turn_spd_skw = skewness(sp(tsi:tei),0); turn_spd_kur = kurtosis(sp(tsi:tei),0); turn_spd_s_t = sum(sp(tsi:tei))/turn_time; turn_spd_s_d = sum(sp(tsi:tei))/turn_dis; turn_com_avg = mean(com(tsi:tei)); turn_com_rms = norm(com(tsi:tei))/sqrt(length(com(tsi:tei))); turn_com_int = sum(com(tsi:tei))*0.1*length(com(tsi:tei)); turn_com_var = var(com(tsi:tei),1); turn_com_med = median(com(tsi:tei)); turn_com_ran = turn_com_max turn_com_min; turn_com_skw = skewness(com(tsi:tei),0); turn_com_kur = kurtosis(com(tsi:tei),0); turn_com_s_t = sum(com(tsi:tei))/turn_time; turn_com_s_d = sum(com(tsi:tei))/turn_dis; % recovery if k == 1 | k == 13 factor = 1; rec_yaw_min = min(ar(rsi:rei)); rec_yaw_max = max(ar(rsi:rei)); rec_lat_min = min(al(rsi:rei)); rec_lat_max = max(al(rsi:rei)); rec_str_min = min(sw(rsi:rei)); rec_str_max = max(sw(rsi:rei)); else factor = -1; rec_yaw_min = -max(ar(rsi:rei)); rec_yaw_max = -min(ar(rsi:rei)); rec_lat_min = -max(al(rsi:rei)); rec_lat_max = -min(al(rsi:rei)); rec_str_min = -max(sw(rsi:rei)); rec_str_max = -min(sw(rsi:rei)); end rec_fwd_min = min(af(rsi:rei)); rec_fwd_max = max(af(rsi:rei)); rec_spd_min = min(sp(rsi:rei)); rec_spd_max = max(sp(rsi:rei)); rec_com_min = min(com(rsi:rei)); rec_com_max = max(com(rsi:rei)); rec_yaw_avg = mean(factor.*ar(rsi:rei)); rec_yaw_rms = norm(factor.*ar(rsi:rei))/sqrt(length(ar(rsi:rei))); rec_yaw_int = sum(factor.*ar(rsi:rei))*0.1*length(ar(rsi:rei)); rec_yaw_var = var(factor.*ar(rsi:rei),1); rec_yaw_med = median(factor.*ar(rsi:rei)); rec_yaw_ran = rec_yaw_max rec_yaw_min; rec_yaw_skw = skewness(factor.*ar(rsi:rei),0); rec_yaw_kur = kurtosis(factor.*ar(rsi:rei),0); rec_yaw_s_t = sum(factor.*ar(rsi:rei))/rec_time; rec_yaw_s_d = sum(factor.*ar(rsi:rei))/rec_dis; rec_lat_avg = mean(factor.*al(rsi:rei)); rec_lat_rms = norm(factor.*al(rsi:rei))/sqrt(length(al(rsi:rei))); rec_lat_int = sum(factor.*al(rsi:rei))*0.1*length(al(rsi:rei)); rec_lat_var = var(factor.*al(rsi:rei),1); rec_lat_med = median(factor.*al(rsi:rei)); rec_lat_ran = rec_lat_max rec_lat_min; rec_lat_skw = skewness(factor.*al(rsi:rei),0); rec_lat_kur = kurtosis(factor.*al(rsi:rei),0); rec_lat_s_t = sum(factor.*al(rsi:rei))/rec_time; rec_lat_s_d = sum(factor.*al(rsi:rei))/rec_dis; rec_fwd_avg = mean(af(rsi:rei)); rec_fwd_rms = norm(af(rsi:rei))/sqrt(length(af(rsi:rei))); rec_fwd_int = sum(af(rsi:rei))*0.1*length(af(rsi:rei)); rec_fwd_var = var(af(rsi:rei),1); rec_fwd_med = median(af(rsi:rei)); rec_fwd_ran = rec_fwd_max rec_fwd_min; rec_fwd_skw = skewness(af(rsi:rei),0); rec_fwd_kur = kurtosis(af(rsi:rei),0); rec_fwd_s_t = sum(af(rsi:rei))/rec_time; rec_fwd_s_d = sum(af(rsi:rei))/rec_dis; rec_str_avg = mean(factor.*sw(rsi:rei)); rec_str_rms = norm(factor.*sw(rsi:rei))/sqrt(length(sw(rsi:rei))); rec_str_int = sum(factor.*sw(rsi:rei))*0.1*length(sw(rsi:rei)); rec_str_var = var(factor.*sw(rsi:rei),1); rec_str_med = median(factor.*sw(rsi:rei)); rec_str_ran = rec_str_max rec_str_min; rec_str_skw = skewness(factor.*sw(rsi:rei),0);

PAGE 118

106 rec_str_kur = kurtosis(factor.*sw(rsi:rei),0); rec_str_s_t = sum(factor.*sw(rsi:rei))/rec_time; rec_str_s_d = sum(factor.*sw(rsi:rei))/rec_dis; rec_spd_avg = mean(sp(rsi:rei)); rec_spd_rms = norm(sp(rsi:rei))/sqrt(length(sp(rsi:rei))); rec_spd_int = sum(sp(rsi:rei))*0.1*length(sp(rsi:rei)); rec_spd_var = var(sp(rsi:rei),1); rec_spd_med = median(sp(rsi:rei)); rec_spd_ran = rec_spd_max rec_spd_min; rec_spd_skw = skewness(sp(rsi:rei),0); rec_spd_kur = kurtosis(sp(rsi:rei),0); rec_spd_s_t = sum(sp(rsi:rei))/rec_time; rec_spd_s_d = sum(sp(rsi:rei))/rec_dis; rec_com_avg = mean(com(rsi:rei)); rec_com_rms = norm(com(rsi:rei))/sqrt(length(com(rsi:rei))); rec_com_int = sum(com(rsi:rei))*0.1*length(com(rsi:rei)); rec_com_var = var(com(rsi:rei),1); rec_com_med = median(com(rsi:rei)); rec_com_ran = rec_com_max rec_com_min; rec_com_skw = skewness(com(rsi:rei),0); rec_com_kur = kurtosis(com(rsi:rei),0); rec_com_s_t = sum(com(rsi:rei))/rec_time; rec_com_s_d = sum(com(rsi:rei))/rec_dis; % compile statistics man = str2num(A(k)); app_temp = [sub,age_1,age_2,gen,man,imp,stop,app_time,app_dis,app_yaw_min,app_yaw_max,app_yaw_avg,ap p_yaw_rms,app_yaw_int,app_yaw_var,app_yaw_med,app_yaw_ran,app_yaw_skw,app_yaw_kur,app_yaw _s_t,app_yaw_s_d,app_lat_min,app_lat_max,app_lat_avg,app_lat_rms,app_lat_int,app_lat_var, app_lat_med,app_lat_ran,app_lat_skw,app_lat_kur,app_lat_s_t,app_lat_s_d,app_fwd_min,app_f wd_max,app_fwd_avg,app_fwd_rms,app_fwd_int,app_fwd_var,app_fwd_med,app_fwd_ran,app_fwd_sk w,app_fwd_kur,app_fwd_s_t,app_fwd_s_d,app_str_min,app_str_max,app_str_avg,app_str_rms,app _str_int,app_str_var,app_str_med,app_str_ran,app_str_skw,app_str_kur,app_str_s_t,app_str_ s_d,app_spd_min,app_spd_max,app_spd_avg,app_spd_rms,app_spd_int,app_spd_var,app_spd_med,a pp_spd_ran,app_spd_skw,app_spd_kur,app_spd_s_t,app_spd_s_d,app_com_min,app_com_max,app_co m_avg,app_com_rms,app_com_int,app_com_var,app_com_med,app_com_ran,app_com_skw,app_com_kur ,app_com_s_t,app_com_s_d]; if length(app) > 1 app = [app;app_temp]; % approach stats else app = [app_temp]; end turn_temp = [sub,age_1,age_2,gen,man,imp,stop,turn_time,turn_dis,turn_yaw_min,turn_yaw_max,turn_yaw_a vg,turn_yaw_rms,turn_yaw_int,turn_yaw_var,turn_yaw_med,turn_yaw_ran,turn_yaw_skw,turn_yaw _kur,turn_yaw_s_t,turn_yaw_s_d,turn_lat_min,turn_lat_max,turn_lat_avg,turn_lat_rms,turn_l at_int,turn_lat_var,turn_lat_med,turn_lat_ran,turn_lat_skw,turn_lat_kur,turn_lat_s_t,turn _lat_s_d,turn_fwd_min,turn_fwd_max,turn_fwd_avg,turn_fwd_rms,turn_fwd_int,turn_fwd_var,tu rn_fwd_med,turn_fwd_ran,turn_fwd_skw,turn_fwd_kur,turn_fwd_s_t,turn_fwd_s_d,turn_str_min, turn_str_max,turn_str_avg,turn_str_rms,turn_str_int,turn_str_var,turn_str_med,turn_str_ra n,turn_str_skw,turn_str_kur,turn_str_s_t,turn_str_s_d,turn_spd_min,turn_spd_max,turn_spd_ avg,turn_spd_rms,turn_spd_int,turn_spd_var,turn_spd_med,turn_spd_ran,turn_spd_skw,turn_sp d_kur,turn_spd_s_t,turn_spd_s_d,turn_com_min,turn_com_max,turn_com_avg,turn_com_rms,turn_ com_int,turn_com_var,turn_com_med,turn_com_ran,turn_com_skw,turn_com_kur,turn_com_s_t,tur n_com_s_d]; if length(turn) > 1 turn = [turn;turn_temp];% turn stats else turn = [turn_temp]; end rec_temp = [sub,age_1,age_2,gen,man,imp,stop,rec_time,rec_dis,rec_yaw_min,rec_yaw_max,rec_yaw_avg,re c_yaw_rms,rec_yaw_int,rec_yaw_var,rec_yaw_med,rec_yaw_ran,rec_yaw_skw,rec_yaw_kur,rec_yaw _s_t,rec_yaw_s_d,rec_lat_min,rec_lat_max,rec_lat_avg,rec_lat_rms,rec_lat_int,rec_lat_var, rec_lat_med,rec_lat_ran,rec_lat_skw,rec_lat_kur,rec_lat_s_t,rec_lat_s_d,rec_fwd_min,rec_f wd_max,rec_fwd_avg,rec_fwd_rms,rec_fwd_int,rec_fwd_var,rec_fwd_med,rec_fwd_ran,rec_fwd_sk w,rec_fwd_kur,rec_fwd_s_t,rec_fwd_s_d,rec_str_min,rec_str_max,rec_str_avg,rec_str_rms,rec _str_int,rec_str_var,rec_str_med,rec_str_ran,rec_str_skw,rec_str_kur,rec_str_s_t,rec_str_ s_d,rec_spd_min,rec_spd_max,rec_spd_avg,rec_spd_rms,rec_spd_int,rec_spd_var,rec_spd_med,r ec_spd_ran,rec_spd_skw,rec_spd_kur,rec_spd_s_t,rec_spd_s_d,rec_com_min,rec_com_max,rec_co m_avg,rec_com_rms,rec_com_int,rec_com_var,rec_com_med,rec_com_ran,rec_com_skw,rec_com_kur ,rec_com_s_t,rec_com_s_d]; if length(rec) > 1 rec = [rec;rec_temp]; % recovery stats else rec = [rec_temp];

PAGE 119

107 end % plot each maneuver figure((k+1)/2);clf plot(t,al,'Color',[0 0 0],'LineStyle','-');hold on plot(t,af,'Color',[0 0 0],'LineStyle',':'); plot(t,sw,'Color',[0.5 1 0],'LineStyle',':'); plot(t,ar,'Color',[0.5 1 0],'LineStyle','-'); plot(t2,sp2/20,'Color',[0 0.5 1],'LineStyle','-.','LineWidth',2); plot(t(tsi),0,'Marker','x','LineWidth',2,'MarkerSize',16,'Color',[0,0,0]); plot(t(tei),0,'Marker','x','LineWidth',2,'MarkerSize',16,'Color',[0,0,0]);hold off st = ['(Subject )'];TitleName2 = [TitleName,' ',st(1:9),num2str(sub),st(10)]; ylim([-1 1.5]) title(TitleName2) legend('LA','FA','ST','YW','SP',0); xlabel('time [sec]') ylabel('g, rev, rad/s, mph/20') end % create statistic output files appstr = ['app_']; subject = ['subject_']; AppName = [appstr,subject,num2str(sub),'.txt']; dlmwrite(AppName,turn) turnstr = ['turn_']; TurnName = [turnstr,subject,num2str(sub),'.txt']; dlmwrite(TurnName,turn) recstr = ['recovery_']; RecName = [recstr,subject,num2str(sub),'.txt']; dlmwrite(RecName,rec) status = 'Finished' sort.m % ********************************************************************** % % FHWA Sorting Program (organizes stats by maneuver) % % University of Florida % College of Publich Health and Health Professions % Occupational Therapy Department % Created by: Ethan J. Davis (Mechanical & Aerospace Engineering) % Date: 09-15-05 % % -------------------------INSTRUCTIONS --------------------------% Place this M-File in the same folder with all of the Statistic Files % (i.e. "turn_subject_23.txt" ) % Open Matlab, Open this M-File % Run this M-File (F5) % % --------------------------------------------------------------------% *** NOTE: In order to process a given subject, the subject's number % must be added to the "subs" vector on line 25 (see below) *** % --------------------------------------------------------------------% % ********************************************************************** clear all subs = [1,2,3]; % <<---ADD SUBJECT NUMBERS HERE app_str = ['app_']; turn_str = ['turn_']; rec_str = ['recovery_']; sub_str = ['subject_']; % initialize arrays man_1_app = 0; man_2_app = 0; man_3_app = 0; man_4_app = 0; man_6_app = 0; man_1_turn = 0; man_2_turn = 0; man_3_turn = 0; man_4_turn = 0; man_6_turn = 0; man_1_rec = 0; man_2_rec = 0;

PAGE 120

108 man_3_rec = 0; man_4_rec = 0; man_6_rec = 0; % process 1 subject at a time for cnt = 1:length(subs) app_file_name = [app_str,sub_str,num2str(subs(cnt)),'.txt']; app_data = dlmread(app_file_name); % read approach file turn_file_name = [turn_str,sub_str,num2str(subs(cnt)),'.txt']; turn_data = dlmread(turn_file_name);% read turn file rec_file_name = [rec_str,sub_str,num2str(subs(cnt)),'.txt']; rec_data = dlmread(rec_file_name); % read recovery file % sort each file by maneuver man_1A_temp_app = app_data(3,:); man_1B_temp_app = app_data(7,:); man_2A_temp_app = app_data(1,:); man_2B_temp_app = app_data(8,:); man_3A_temp_app = app_data(10,:); man_3B_temp_app = app_data(9,:); man_4A_temp_app = app_data(5,:); man_4B_temp_app = app_data(4,:); man_6A_temp_app = app_data(6,:); man_6B_temp_app = app_data(2,:); man_1A_temp_turn = turn_data(3,:); man_1B_temp_turn = turn_data(7,:); man_2A_temp_turn = turn_data(1,:); man_2B_temp_turn = turn_data(8,:); man_3A_temp_turn = turn_data(10,:); man_3B_temp_turn = turn_data(9,:); man_4A_temp_turn = turn_data(5,:); man_4B_temp_turn = turn_data(4,:); man_6A_temp_turn = turn_data(6,:); man_6B_temp_turn = turn_data(2,:); man_1A_temp_rec = rec_data(3,:); man_1B_temp_rec = rec_data(7,:); man_2A_temp_rec = rec_data(1,:); man_2B_temp_rec = rec_data(8,:); man_3A_temp_rec = rec_data(10,:); man_3B_temp_rec = rec_data(9,:); man_4A_temp_rec = rec_data(5,:); man_4B_temp_rec = rec_data(4,:); man_6A_temp_rec = rec_data(6,:); man_6B_temp_rec = rec_data(2,:); % create statistic files compiled by maneuver if length(man_1_turn) > 1 man_1_app = [man_1_app;man_1A_temp_app,man_1B_temp_app]; man_2_app = [man_2_app;man_2A_temp_app,man_2B_temp_app]; man_3_app = [man_3_app;man_3A_temp_app,man_3B_temp_app]; man_4_app = [man_4_app;man_4A_temp_app,man_4B_temp_app]; man_6_app = [man_6_app;man_6A_temp_app,man_6B_temp_app]; man_1_turn = [man_1_turn;man_1A_temp_turn,man_1B_temp_turn]; man_2_turn = [man_2_turn;man_2A_temp_turn,man_2B_temp_turn]; man_3_turn = [man_3_turn;man_3A_temp_turn,man_3B_temp_turn]; man_4_turn = [man_4_turn;man_4A_temp_turn,man_4B_temp_turn]; man_6_turn = [man_6_turn;man_6A_temp_turn,man_6B_temp_turn]; man_1_rec = [man_1_rec;man_1A_temp_rec,man_1B_temp_rec]; man_2_rec = [man_2_rec;man_2A_temp_rec,man_2B_temp_rec]; man_3_rec = [man_3_rec;man_3A_temp_rec,man_3B_temp_rec]; man_4_rec = [man_4_rec;man_4A_temp_rec,man_4B_temp_rec]; man_6_rec = [man_6_rec;man_6A_temp_rec,man_6B_temp_rec]; else man_1_app = [man_1A_temp_app,man_1B_temp_app]; man_2_app = [man_2A_temp_app,man_2B_temp_app]; man_3_app = [man_3A_temp_app,man_3B_temp_app]; man_4_app = [man_4A_temp_app,man_4B_temp_app]; man_6_app = [man_6A_temp_app,man_6B_temp_app]; man_1_turn = [man_1A_temp_turn,man_1B_temp_turn]; man_2_turn = [man_2A_temp_turn,man_2B_temp_turn]; man_3_turn = [man_3A_temp_turn,man_3B_temp_turn]; man_4_turn = [man_4A_temp_turn,man_4B_temp_turn]; man_6_turn = [man_6A_temp_turn,man_6B_temp_turn]; man_1_rec = [man_1A_temp_rec,man_1B_temp_rec]; man_2_rec = [man_2A_temp_rec,man_2B_temp_rec]; man_3_rec = [man_3A_temp_rec,man_3B_temp_rec]; man_4_rec = [man_4A_temp_rec,man_4B_temp_rec]; man_6_rec = [man_6A_temp_rec,man_6B_temp_rec]; end

PAGE 121

109 end man_str = ['Maneuver_']; sta_str = ['_Stats']; app_str = ['_Approach']; turn_str = ['_Turn']; rec_str = ['_Recovery']; % create output files man_1_fn_app = [man_str,num2str(1),app_str,sta_str,'.txt']; man_2_fn_app = [man_str,num2str(2),app_str,sta_str,'.txt']; man_3_fn_app = [man_str,num2str(3),app_str,sta_str,'.txt']; man_4_fn_app = [man_str,num2str(4),app_str,sta_str,'.txt']; man_6_fn_app = [man_str,num2str(6),app_str,sta_str,'.txt']; dlmwrite(man_1_fn_app,man_1_app) dlmwrite(man_2_fn_app,man_2_app) dlmwrite(man_3_fn_app,man_3_app) dlmwrite(man_4_fn_app,man_4_app) dlmwrite(man_6_fn_app,man_6_app) man_1_fn_turn = [man_str,num2str(1),turn_str,sta_str,'.txt']; man_2_fn_turn = [man_str,num2str(2),turn_str,sta_str,'.txt']; man_3_fn_turn = [man_str,num2str(3),turn_str,sta_str,'.txt']; man_4_fn_turn = [man_str,num2str(4),turn_str,sta_str,'.txt']; man_6_fn_turn = [man_str,num2str(6),turn_str,sta_str,'.txt']; dlmwrite(man_1_fn_turn,man_1_turn) dlmwrite(man_2_fn_turn,man_2_turn) dlmwrite(man_3_fn_turn,man_3_turn) dlmwrite(man_4_fn_turn,man_4_turn) dlmwrite(man_6_fn_turn,man_6_turn) man_1_fn_rec = [man_str,num2str(1),rec_str,sta_str,'.txt']; man_2_fn_rec = [man_str,num2str(2),rec_str,sta_str,'.txt']; man_3_fn_rec = [man_str,num2str(3),rec_str,sta_str,'.txt']; man_4_fn_rec = [man_str,num2str(4),rec_str,sta_str,'.txt']; man_6_fn_rec = [man_str,num2str(6),rec_str,sta_str,'.txt']; dlmwrite(man_1_fn_rec,man_1_rec) dlmwrite(man_2_fn_rec,man_2_rec) dlmwrite(man_3_fn_rec,man_3_rec) dlmwrite(man_4_fn_rec,man_4_rec) dlmwrite(man_6_fn_rec,man_6_rec) status = 'Finished' sim.m % ********************************************************************** % % Simulator Scenario Analysis % % University of Florida % College of Publich Health and Health Professions % Occupational Therapy Department % Created by: Ethan J. Davis (Mechanical & Aerospace Engineering) % Date: 8-25-05 % % -------------------------INSTRUCTIONS --------------------------% Place this M-File in the same folder with all of the Scenario % Files (i.e. "scenario_1a.csv" ) % Open Matlab, Open this M-File % Run this M-File (F5) % % ********************************************************************** clear all format short g sub = input('What is the Subject Number ? '); age = input('How old is the subject? '); gen = input('What gender is the subject? (0 = MALE, 1 = FEMALE) '); % determine age groups if age < 65 age_1 = 0; age_2 = 0; elseif age > 64 & age < 76 age_1 = 1; age_2 = 1; elseif age > 75 & age < 86 age_1 = 2; age_2 = 1;

PAGE 122

110 else age_1 = 3; age_2 = 1; end man_labels = ['1a1b1c2a2b3a3b4a4b5a5b']; % maneuver labels str_1 = ['fhwa']; str_2 = ['scenario_']; turn = 0; rec = 0; app = 0; % process one maneuver at a time for k = 1:2:21 filename = [str_2,man_labels(k:k+1),'.csv']; data = dlmread(filename); % read file % convert improvement letters to numbers if man_labels(k+1) == 'a' imp = 1; elseif man_labels(k+1) == 'b' imp = 2; elseif man_labels(k+1) == 'c' imp = 3; else status = 'ERROR' end af = data(:,2)/32.174; % forward accel al = data(:,3)/32.174; % lateral accel sw = data(:,7)/360; % steering ar = data(:,9); % yaw sp = data(:,8); % speed head = data(:,6); % heading t = 0.1:0.1:length(head)/10;% time if k == 7 | k == 9 % right turns f = find(head > 45); % find middle of turn else f = find(head < -45); % find middle of turn end f2 = find(abs(ar) < 0.05); f3 = find(f2 < min(f)); f4 = find(f2 > min(f)); tsi = f2(max(f3)); % turn start index tei = f2(min(f4)); % turn end index rsi = tei + 1; % recovery start index rei = length(af); % recovery end index asi = 1; % approach start index aei = tsi 1; % approach end index % plot maneuver figure((k+1)/2);clf plot(t,al(rangelo:rangehi),'Color',[0 0 0],'LineStyle','-');hold on plot(t,af(rangelo:rangehi),'Color',[0 0 0],'LineStyle',':'); plot(t,sw(rangelo:rangehi),'Color',[0.5 1 0],'LineStyle',':'); plot(t,ar(rangelo:rangehi),'Color',[0.5 1 0],'LineStyle','-'); plot(t,sp/20,'Color',[0 0.5 1],'LineStyle','-.','LineWidth',2); plot(t(tsi),0,'Marker','x','MarkerSize',12,'Color',[0,0,0]); plot(t(tei),0,'Marker','x','MarkerSize',12,'Color',[0,0,0]);hold ylim([-2 2]) legend('LA','FA','ST','YW','SP',0) st = ['(Subject )']; title_str = ['Maneuver ',man_labels(k:k+1),' ',st(1:9),num2str(sub),st(10)]; title(title_str) xlabel('Time [sec]') ylabel('[g], [rev], [rad/s], [mph/20]') f_stop = find(sp == 0); % determine if subject stopped if length(f_stop) > 0 stop = 1; else stop = 0; end turn_dis = 1; rec_dis = 1; app_dis = 1; app_time = t(aei) t(asi); % time to complete approach phase turn_time = t(tei) t(tsi); % time to complete turning phase rec_time = t(rei) t(rsi); % time to complete recovery phase com = 0;

PAGE 123

111 for i = 1:length(al) com(i) = sqrt(af(i)^2+al(i)^2); % combined acceleration for turn end % calculate statistics for each measurement of each phase % approach if k == 7 | k == 9 factor = 1; app_yaw_min = min(ar(asi:aei)); % minumum app_yaw_max = max(ar(asi:aei)); % maximum app_lat_min = min(al(asi:aei)); app_lat_max = max(al(asi:aei)); app_str_min = min(sw(asi:aei)); app_str_max = max(sw(asi:aei)); else factor = -1; app_yaw_min = -max(ar(asi:aei)); app_yaw_max = -min(ar(asi:aei)); app_lat_min = -max(al(asi:aei)); app_lat_max = -min(al(asi:aei)); app_str_min = -max(sw(asi:aei)); app_str_max = -min(sw(asi:aei)); end app_fwd_min = min(af(asi:aei)); app_fwd_max = max(af(asi:aei)); app_spd_min = min(sp(asi:aei)); app_spd_max = max(sp(asi:aei)); app_com_min = min(com(asi:aei)); app_com_max = max(com(asi:aei)); app_yaw_avg = mean(factor.*ar(asi:aei)); % mean app_yaw_rms = norm(factor.*ar(asi:aei))/sqrt(length(ar(asi:aei))); % rms app_yaw_int = sum(factor.*ar(asi:aei))*0.1*length(ar(asi:aei)); % integral app_yaw_var = var(factor.*ar(asi:aei),1); % variance app_yaw_med = median(factor.*ar(asi:aei)); % median app_yaw_ran = app_yaw_max app_yaw_min; % range app_yaw_skw = skewness(factor.*ar(asi:aei),0); % skewness app_yaw_kur = kurtosis(factor.*ar(asi:aei),0); % kurtosis app_yaw_s_t = sum(factor.*ar(asi:aei))/app_time; % sum/time app_yaw_s_d = sum(factor.*ar(asi:aei))/app_dis; % sum/distance app_lat_avg = mean(factor.*al(asi:aei)); app_lat_rms = norm(factor.*al(asi:aei))/sqrt(length(al(asi:aei))); app_lat_int = sum(factor.*al(asi:aei))*0.1*length(al(asi:aei)); app_lat_var = var(factor.*al(asi:aei),1); app_lat_med = median(factor.*al(asi:aei)); app_lat_ran = app_lat_max app_lat_min; app_lat_skw = skewness(factor.*al(asi:aei),0); app_lat_kur = kurtosis(factor.*al(asi:aei),0); app_lat_s_t = sum(factor.*al(asi:aei))/app_time; app_lat_s_d = sum(factor.*al(asi:aei))/app_dis; app_fwd_avg = mean(af(asi:aei)); app_fwd_rms = norm(af(asi:aei))/sqrt(length(af(asi:aei))); app_fwd_int = sum(af(asi:aei))*0.1*length(af(asi:aei)); app_fwd_var = var(af(asi:aei),1); app_fwd_med = median(af(asi:aei)); app_fwd_ran = app_fwd_max app_fwd_min; app_fwd_skw = skewness(af(asi:aei),0); app_fwd_kur = kurtosis(af(asi:aei),0); app_fwd_s_t = sum(af(asi:aei))/app_time; app_fwd_s_d = sum(af(asi:aei))/app_dis; app_str_avg = mean(factor.*sw(asi:aei)); app_str_rms = norm(factor.*sw(asi:aei))/sqrt(length(sw(asi:aei))); app_str_int = sum(factor.*sw(asi:aei))*0.1*length(sw(asi:aei)); app_str_var = var(factor.*sw(asi:aei),1); app_str_med = median(factor.*sw(asi:aei)); app_str_ran = app_str_max app_str_min; app_str_skw = skewness(factor.*sw(asi:aei),0); app_str_kur = kurtosis(factor.*sw(asi:aei),0); app_str_s_t = sum(factor.*sw(asi:aei))/app_time; app_str_s_d = sum(factor.*sw(asi:aei))/app_dis; app_spd_avg = mean(sp(asi:aei)); app_spd_rms = norm(sp(asi:aei))/sqrt(length(sp(asi:aei))); app_spd_int = sum(sp(asi:aei))*0.1*length(sp(asi:aei)); app_spd_var = var(sp(asi:aei),1); app_spd_med = median(sp(asi:aei)); app_spd_ran = app_spd_max app_spd_min; app_spd_skw = skewness(sp(asi:aei),0);

PAGE 124

112 app_spd_kur = kurtosis(sp(asi:aei),0); app_spd_s_t = sum(sp(asi:aei))/app_time; app_spd_s_d = sum(sp(asi:aei))/app_dis; app_com_avg = mean(com(asi:aei)); app_com_rms = norm(com(asi:aei))/sqrt(length(com(asi:aei))); app_com_int = sum(com(asi:aei))*0.1*length(com(asi:aei)); app_com_var = var(com(asi:aei),1); app_com_med = median(com(asi:aei)); app_com_ran = app_com_max app_com_min; app_com_skw = skewness(com(asi:aei),0); app_com_kur = kurtosis(com(asi:aei),0); app_com_s_t = sum(com(asi:aei))/app_time; app_com_s_d = sum(com(asi:aei))/app_dis; % turn if k == 7 | k == 9 factor = 1; turn_yaw_min = min(ar(tsi:tei)); turn_yaw_max = max(ar(tsi:tei)); turn_lat_min = min(al(tsi:tei)); turn_lat_max = max(al(tsi:tei)); turn_str_min = min(sw(tsi:tei)); turn_str_max = max(sw(tsi:tei)); else factor = -1; turn_yaw_min = -max(ar(tsi:tei)); turn_yaw_max = -min(ar(tsi:tei)); turn_lat_min = -max(al(tsi:tei)); turn_lat_max = -min(al(tsi:tei)); turn_str_min = -max(sw(tsi:tei)); turn_str_max = -min(sw(tsi:tei)); end turn_fwd_min = min(af(tsi:tei)); turn_fwd_max = max(af(tsi:tei)); turn_spd_min = min(sp(tsi:tei)); turn_spd_max = max(sp(tsi:tei)); turn_com_min = min(com(tsi:tei)); turn_com_max = max(com(tsi:tei)); turn_yaw_avg = mean(factor.*ar(tsi:tei)); turn_yaw_rms = norm(factor.*ar(tsi:tei))/sqrt(length(ar(tsi:tei))); turn_yaw_int = sum(factor.*ar(tsi:tei))*0.1*length(ar(tsi:tei)); turn_yaw_var = var(factor.*ar(tsi:tei),1); turn_yaw_med = median(factor.*ar(tsi:tei)); turn_yaw_ran = turn_yaw_max turn_yaw_min; turn_yaw_skw = skewness(factor.*ar(tsi:tei),0); turn_yaw_kur = kurtosis(factor.*ar(tsi:tei),0); turn_yaw_s_t = sum(factor.*ar(tsi:tei))/turn_time; turn_yaw_s_d = sum(factor.*ar(tsi:tei))/turn_dis; turn_lat_avg = mean(factor.*al(tsi:tei)); turn_lat_rms = norm(factor.*al(tsi:tei))/sqrt(length(al(tsi:tei))); turn_lat_int = sum(factor.*al(tsi:tei))*0.1*length(al(tsi:tei)); turn_lat_var = var(factor.*al(tsi:tei),1); turn_lat_med = median(factor.*al(tsi:tei)); turn_lat_ran = turn_lat_max turn_lat_min; turn_lat_skw = skewness(factor.*al(tsi:tei),0); turn_lat_kur = kurtosis(factor.*al(tsi:tei),0); turn_lat_s_t = sum(factor.*al(tsi:tei))/turn_time; turn_lat_s_d = sum(factor.*al(tsi:tei))/turn_dis; turn_fwd_avg = mean(af(tsi:tei)); turn_fwd_rms = norm(af(tsi:tei))/sqrt(length(af(tsi:tei))); turn_fwd_int = sum(af(tsi:tei))*0.1*length(af(tsi:tei)); turn_fwd_var = var(af(tsi:tei),1); turn_fwd_med = median(af(tsi:tei)); turn_fwd_ran = turn_fwd_max turn_fwd_min; turn_fwd_skw = skewness(af(tsi:tei),0); turn_fwd_kur = kurtosis(af(tsi:tei),0); turn_fwd_s_t = sum(af(tsi:tei))/turn_time; turn_fwd_s_d = sum(af(tsi:tei))/turn_dis; turn_str_avg = mean(factor.*sw(tsi:tei)); turn_str_rms = norm(factor.*sw(tsi:tei))/sqrt(length(sw(tsi:tei))); turn_str_int = sum(factor.*sw(tsi:tei))*0.1*length(sw(tsi:tei)); turn_str_var = var(factor.*sw(tsi:tei),1); turn_str_med = median(factor.*sw(tsi:tei)); turn_str_ran = turn_str_max turn_str_min; turn_str_skw = skewness(factor.*sw(tsi:tei),0); turn_str_kur = kurtosis(factor.*sw(tsi:tei),0);

PAGE 125

113 turn_str_s_t = sum(factor.*sw(tsi:tei))/turn_time; turn_str_s_d = sum(factor.*sw(tsi:tei))/turn_dis; turn_spd_avg = mean(sp(tsi:tei)); turn_spd_rms = norm(sp(tsi:tei))/sqrt(length(sp(tsi:tei))); turn_spd_int = sum(sp(tsi:tei))*0.1*length(sp(tsi:tei)); turn_spd_var = var(sp(tsi:tei),1); turn_spd_med = median(sp(tsi:tei)); turn_spd_ran = turn_spd_max turn_spd_min; turn_spd_skw = skewness(sp(tsi:tei),0); turn_spd_kur = kurtosis(sp(tsi:tei),0); turn_spd_s_t = sum(sp(tsi:tei))/turn_time; turn_spd_s_d = sum(sp(tsi:tei))/turn_dis; turn_com_avg = mean(com(tsi:tei)); turn_com_rms = norm(com(tsi:tei))/sqrt(length(com(tsi:tei))); turn_com_int = sum(com(tsi:tei))*0.1*length(com(tsi:tei)); turn_com_var = var(com(tsi:tei),1); turn_com_med = median(com(tsi:tei)); turn_com_ran = turn_com_max turn_com_min; turn_com_skw = skewness(com(tsi:tei),0); turn_com_kur = kurtosis(com(tsi:tei),0); turn_com_s_t = sum(com(tsi:tei))/turn_time; turn_com_s_d = sum(com(tsi:tei))/turn_dis; % recovery if k == 7 | k == 9 factor = 1; rec_yaw_min = min(ar(rsi:rei)); rec_yaw_max = max(ar(rsi:rei)); rec_lat_min = min(al(rsi:rei)); rec_lat_max = max(al(rsi:rei)); rec_str_min = min(sw(rsi:rei)); rec_str_max = max(sw(rsi:rei)); else factor = -1; rec_yaw_min = -max(ar(rsi:rei)); rec_yaw_max = -min(ar(rsi:rei)); rec_lat_min = -max(al(rsi:rei)); rec_lat_max = -min(al(rsi:rei)); rec_str_min = -max(sw(rsi:rei)); rec_str_max = -min(sw(rsi:rei)); end rec_fwd_min = min(af(rsi:rei)); rec_fwd_max = max(af(rsi:rei)); rec_spd_min = min(sp(rsi:rei)); rec_spd_max = max(sp(rsi:rei)); rec_com_min = min(com(rsi:rei)); rec_com_max = max(com(rsi:rei)); rec_yaw_avg = mean(factor.*ar(rsi:rei)); rec_yaw_rms = norm(factor.*ar(rsi:rei))/sqrt(length(ar(rsi:rei))); rec_yaw_int = sum(factor.*ar(rsi:rei))*0.1*length(ar(rsi:rei)); rec_yaw_var = var(factor.*ar(rsi:rei),1); rec_yaw_med = median(factor.*ar(rsi:rei)); rec_yaw_ran = rec_yaw_max rec_yaw_min; rec_yaw_skw = skewness(factor.*ar(rsi:rei),0); rec_yaw_kur = kurtosis(factor.*ar(rsi:rei),0); rec_yaw_s_t = sum(factor.*ar(rsi:rei))/rec_time; rec_yaw_s_d = sum(factor.*ar(rsi:rei))/rec_dis; rec_lat_avg = mean(factor.*al(rsi:rei)); rec_lat_rms = norm(factor.*al(rsi:rei))/sqrt(length(al(rsi:rei))); rec_lat_int = sum(factor.*al(rsi:rei))*0.1*length(al(rsi:rei)); rec_lat_var = var(factor.*al(rsi:rei),1); rec_lat_med = median(factor.*al(rsi:rei)); rec_lat_ran = rec_lat_max rec_lat_min; rec_lat_skw = skewness(factor.*al(rsi:rei),0); rec_lat_kur = kurtosis(factor.*al(rsi:rei),0); rec_lat_s_t = sum(factor.*al(rsi:rei))/rec_time; rec_lat_s_d = sum(factor.*al(rsi:rei))/rec_dis; rec_fwd_avg = mean(af(rsi:rei)); rec_fwd_rms = norm(af(rsi:rei))/sqrt(length(af(rsi:rei))); rec_fwd_int = sum(af(rsi:rei))*0.1*length(af(rsi:rei)); rec_fwd_var = var(af(rsi:rei),1); rec_fwd_med = median(af(rsi:rei)); rec_fwd_ran = rec_fwd_max rec_fwd_min; rec_fwd_skw = skewness(af(rsi:rei),0); rec_fwd_kur = kurtosis(af(rsi:rei),0); rec_fwd_s_t = sum(af(rsi:rei))/rec_time;

PAGE 126

114 rec_fwd_s_d = sum(af(rsi:rei))/rec_dis; rec_str_avg = mean(factor.*sw(rsi:rei)); rec_str_rms = norm(factor.*sw(rsi:rei))/sqrt(length(sw(rsi:rei))); rec_str_int = sum(factor.*sw(rsi:rei))*0.1*length(sw(rsi:rei)); rec_str_var = var(factor.*sw(rsi:rei),1); rec_str_med = median(factor.*sw(rsi:rei)); rec_str_ran = rec_str_max rec_str_min; rec_str_skw = skewness(factor.*sw(rsi:rei),0); rec_str_kur = kurtosis(factor.*sw(rsi:rei),0); rec_str_s_t = sum(factor.*sw(rsi:rei))/rec_time; rec_str_s_d = sum(factor.*sw(rsi:rei))/rec_dis; rec_spd_avg = mean(sp(rsi:rei)); rec_spd_rms = norm(sp(rsi:rei))/sqrt(length(sp(rsi:rei))); rec_spd_int = sum(sp(rsi:rei))*0.1*length(sp(rsi:rei)); rec_spd_var = var(sp(rsi:rei),1); rec_spd_med = median(sp(rsi:rei)); rec_spd_ran = rec_spd_max rec_spd_min; rec_spd_skw = skewness(sp(rsi:rei),0); rec_spd_kur = kurtosis(sp(rsi:rei),0); rec_spd_s_t = sum(sp(rsi:rei))/rec_time; rec_spd_s_d = sum(sp(rsi:rei))/rec_dis; rec_com_avg = mean(com(rsi:rei)); rec_com_rms = norm(com(rsi:rei))/sqrt(length(com(rsi:rei))); rec_com_int = sum(com(rsi:rei))*0.1*length(com(rsi:rei)); rec_com_var = var(com(rsi:rei),1); rec_com_med = median(com(rsi:rei)); rec_com_ran = rec_com_max rec_com_min; rec_com_skw = skewness(com(rsi:rei),0); rec_com_kur = kurtosis(com(rsi:rei),0); rec_com_s_t = sum(com(rsi:rei))/rec_time; rec_com_s_d = sum(com(rsi:rei))/rec_dis; % compile statistics by phase (contains all maneuvers) man = str2num(man_labels(k)); app_temp = [sub,age_1,age_2,gen,man,imp,stop,app_time,app_dis,app_yaw_min,app_yaw_max,app_yaw_avg,ap p_yaw_rms,app_yaw_int,app_yaw_var,app_yaw_med,app_yaw_ran,app_yaw_skw,app_yaw_kur,app_yaw _s_t,app_yaw_s_d,app_lat_min,app_lat_max,app_lat_avg,app_lat_rms,app_lat_int,app_lat_var, app_lat_med,app_lat_ran,app_lat_skw,app_lat_kur,app_lat_s_t,app_lat_s_d,app_fwd_min,app_f wd_max,app_fwd_avg,app_fwd_rms,app_fwd_int,app_fwd_var,app_fwd_med,app_fwd_ran,app_fwd_sk w,app_fwd_kur,app_fwd_s_t,app_fwd_s_d,app_str_min,app_str_max,app_str_avg,app_str_rms,app _str_int,app_str_var,app_str_med,app_str_ran,app_str_skw,app_str_kur,app_str_s_t,app_str_ s_d,app_spd_min,app_spd_max,app_spd_avg,app_spd_rms,app_spd_int,app_spd_var,app_spd_med,a pp_spd_ran,app_spd_skw,app_spd_kur,app_spd_s_t,app_spd_s_d,app_com_min,app_com_max,app_co m_avg,app_com_rms,app_com_int,app_com_var,app_com_med,app_com_ran,app_com_skw,app_com_kur ,app_com_s_t,app_com_s_d]; if length(app) > 1 app = [app;app_temp]; % approach stats else app = [app_temp]; end turn_temp = [sub,age_1,age_2,gen,man,imp,stop,turn_time,turn_dis,turn_yaw_min,turn_yaw_max,turn_yaw_a vg,turn_yaw_rms,turn_yaw_int,turn_yaw_var,turn_yaw_med,turn_yaw_ran,turn_yaw_skw,turn_yaw _kur,turn_yaw_s_t,turn_yaw_s_d,turn_lat_min,turn_lat_max,turn_lat_avg,turn_lat_rms,turn_l at_int,turn_lat_var,turn_lat_med,turn_lat_ran,turn_lat_skw,turn_lat_kur,turn_lat_s_t,turn _lat_s_d,turn_fwd_min,turn_fwd_max,turn_fwd_avg,turn_fwd_rms,turn_fwd_int,turn_fwd_var,tu rn_fwd_med,turn_fwd_ran,turn_fwd_skw,turn_fwd_kur,turn_fwd_s_t,turn_fwd_s_d,turn_str_min, turn_str_max,turn_str_avg,turn_str_rms,turn_str_int,turn_str_var,turn_str_med,turn_str_ra n,turn_str_skw,turn_str_kur,turn_str_s_t,turn_str_s_d,turn_spd_min,turn_spd_max,turn_spd_ avg,turn_spd_rms,turn_spd_int,turn_spd_var,turn_spd_med,turn_spd_ran,turn_spd_skw,turn_sp d_kur,turn_spd_s_t,turn_spd_s_d,turn_com_min,turn_com_max,turn_com_avg,turn_com_rms,turn_ com_int,turn_com_var,turn_com_med,turn_com_ran,turn_com_skw,turn_com_kur,turn_com_s_t,tur n_com_s_d]; if length(turn) > 1 turn = [turn;turn_temp];% turn stats else turn = [turn_temp]; end rec_temp = [sub,age_1,age_2,gen,man,imp,stop,rec_time,rec_dis,rec_yaw_min,rec_yaw_max,rec_yaw_avg,re c_yaw_rms,rec_yaw_int,rec_yaw_var,rec_yaw_med,rec_yaw_ran,rec_yaw_skw,rec_yaw_kur,rec_yaw _s_t,rec_yaw_s_d,rec_lat_min,rec_lat_max,rec_lat_avg,rec_lat_rms,rec_lat_int,rec_lat_var, rec_lat_med,rec_lat_ran,rec_lat_skw,rec_lat_kur,rec_lat_s_t,rec_lat_s_d,rec_fwd_min,rec_f wd_max,rec_fwd_avg,rec_fwd_rms,rec_fwd_int,rec_fwd_var,rec_fwd_med,rec_fwd_ran,rec_fwd_sk w,rec_fwd_kur,rec_fwd_s_t,rec_fwd_s_d,rec_str_min,rec_str_max,rec_str_avg,rec_str_rms,rec _str_int,rec_str_var,rec_str_med,rec_str_ran,rec_str_skw,rec_str_kur,rec_str_s_t,rec_str_ s_d,rec_spd_min,rec_spd_max,rec_spd_avg,rec_spd_rms,rec_spd_int,rec_spd_var,rec_spd_med,r

PAGE 127

115 ec_spd_ran,rec_spd_skw,rec_spd_kur,rec_spd_s_t,rec_spd_s_d,rec_com_min,rec_com_max,rec_co m_avg,rec_com_rms,rec_com_int,rec_com_var,rec_com_med,rec_com_ran,rec_com_skw,rec_com_kur ,rec_com_s_t,rec_com_s_d]; if length(rec) > 1 rec = [rec;rec_temp]; % recovery stats else rec = [rec_temp]; end end % write statisitics files for each phase (contains all maneuvers) subject = ['subject_']; appstr = ['approach_']; AppName = ['SIM_',appstr,subject,num2str(sub),'.txt']; dlmwrite(AppName,turn) turnstr = ['turn_']; TurnName = ['SIM_',turnstr,subject,num2str(sub),'.txt']; dlmwrite(TurnName,turn) recstr = ['recovery_']; RecName = ['SIM_',recstr,subject,num2str(sub),'.txt']; dlmwrite(RecName,rec) status = 'Finished' sim_sort.m % ********************************************************************** % % Simulator Sorting Program (organizes stats by maneuver) % % University of Florida % College of Publich Health and Health Professions % Occupational Therapy Department % Created by: Ethan J. Davis (Mechanical & Aerospace Engineering) % Date: 09-15-05 % % ------------------------INSTRUCTIONS ---------------------------% Place this M-File in the same folder with all of the Statistic Files % (i.e. "SIM_turn_subject_23.txt" ) % Open Matlab, Open this M-File % Run this M-File (F5) % --------------------------------------------------------------------% *** NOTE: In order to process a given subject, the subject's number % must be added to the "subs" vector on line 25 (see below) *** % --------------------------------------------------------------------% % ********************************************************************** clear all status = 'Please Wait...' subs = []; % <-ADD SUBJECT NUMBERS HERE sim_str = ['SIM_']; app_str = ['approach_']; turn_str = ['turn_']; rec_str = ['recovery_']; sub_str = ['subject_']; man_1_app = 0; man_2_app = 0; man_3_app = 0; man_4_app = 0; man_5_app = 0; man_1_turn = 0; man_2_turn = 0; man_3_turn = 0; man_4_turn = 0; man_5_turn = 0; man_1_rec = 0; man_2_rec = 0; man_3_rec = 0; man_4_rec = 0; man_5_rec = 0; % process 1 subject at a time for cnt = 1:length(subs) app_file_name = [sim_str,app_str,sub_str,num2str(subs(cnt)),'.txt']; app_data = dlmread(app_file_name); % read approach file

PAGE 128

116 turn_file_name = [sim_str,turn_str,sub_str,num2str(subs(cnt)),'.txt']; turn_data = dlmread(turn_file_name);% read turn file rec_file_name = [sim_str,rec_str,sub_str,num2str(subs(cnt)),'.txt']; rec_data = dlmread(rec_file_name); % read recovery file % sort by maneuver man_1A_temp_app = app_data(1,:); man_1B_temp_app = app_data(2,:); man_1C_temp_app = app_data(3,:); man_2A_temp_app = app_data(4,:); man_2B_temp_app = app_data(5,:); man_3A_temp_app = app_data(6,:); man_3B_temp_app = app_data(7,:); man_4A_temp_app = app_data(8,:); man_4B_temp_app = app_data(9,:); man_5A_temp_app = app_data(10,:); man_5B_temp_app = app_data(11,:); man_1A_temp_turn = turn_data(1,:); man_1B_temp_turn = turn_data(2,:); man_1C_temp_turn = turn_data(3,:); man_2A_temp_turn = turn_data(4,:); man_2B_temp_turn = turn_data(5,:); man_3A_temp_turn = turn_data(6,:); man_3B_temp_turn = turn_data(7,:); man_4A_temp_turn = turn_data(8,:); man_4B_temp_turn = turn_data(9,:); man_5A_temp_turn = turn_data(10,:); man_5B_temp_turn = turn_data(11,:); man_1A_temp_rec = rec_data(1,:); man_1B_temp_rec = rec_data(2,:); man_1C_temp_rec = rec_data(3,:); man_2A_temp_rec = rec_data(4,:); man_2B_temp_rec = rec_data(5,:); man_3A_temp_rec = rec_data(6,:); man_3B_temp_rec = rec_data(7,:); man_4A_temp_rec = rec_data(8,:); man_4B_temp_rec = rec_data(9,:); man_5A_temp_rec = rec_data(10,:); man_5B_temp_rec = rec_data(11,:); % create statistic arrays if length(man_1_turn) > 1 man_1_app = [man_1_app;man_1A_temp_app,man_1B_temp_app,man_1C_temp_app]; man_2_app = [man_2_app;man_2A_temp_app,man_2B_temp_app]; man_3_app = [man_3_app;man_3A_temp_app,man_3B_temp_app]; man_4_app = [man_4_app;man_4A_temp_app,man_4B_temp_app]; man_5_app = [man_5_app;man_5A_temp_turn,man_5B_temp_turn]; man_1_turn = [man_1_turn;man_1A_temp_turn,man_1B_temp_turn,man_1C_temp_turn]; man_2_turn = [man_2_turn;man_2A_temp_turn,man_2B_temp_turn]; man_3_turn = [man_3_turn;man_3A_temp_turn,man_3B_temp_turn]; man_4_turn = [man_4_turn;man_4A_temp_turn,man_4B_temp_turn]; man_5_turn = [man_5_turn;man_5A_temp_turn,man_5B_temp_turn]; man_1_rec = [man_1_rec;man_1A_temp_rec,man_1B_temp_rec,man_1C_temp_rec]; man_2_rec = [man_2_rec;man_2A_temp_rec,man_2B_temp_rec]; man_3_rec = [man_3_rec;man_3A_temp_rec,man_3B_temp_rec]; man_4_rec = [man_4_rec;man_4A_temp_rec,man_4B_temp_rec]; man_5_rec = [man_5_rec;man_5A_temp_rec,man_5B_temp_rec]; else man_1_app = [man_1A_temp_app,man_1B_temp_app,man_1C_temp_app]; man_2_app = [man_2A_temp_app,man_2B_temp_app]; man_3_app = [man_3A_temp_app,man_3B_temp_app]; man_4_app = [man_4A_temp_app,man_4B_temp_app]; man_5_app = [man_5A_temp_app,man_5B_temp_app]; man_1_turn = [man_1A_temp_turn,man_1B_temp_turn,man_1C_temp_turn]; man_2_turn = [man_2A_temp_turn,man_2B_temp_turn]; man_3_turn = [man_3A_temp_turn,man_3B_temp_turn]; man_4_turn = [man_4A_temp_turn,man_4B_temp_turn]; man_5_turn = [man_5A_temp_turn,man_5B_temp_turn]; man_1_rec = [man_1A_temp_rec,man_1B_temp_rec,man_1C_temp_rec]; man_2_rec = [man_2A_temp_rec,man_2B_temp_rec]; man_3_rec = [man_3A_temp_rec,man_3B_temp_rec]; man_4_rec = [man_4A_temp_rec,man_4B_temp_rec]; man_5_rec = [man_5A_temp_rec,man_5B_temp_rec]; end end man_str = ['Maneuver_']; sta_str = ['_Stats']; app_str = ['_Approach']; turn_str = ['_Turn']; rec_str = ['_Recovery'];

PAGE 129

117 % create output files man_1_fn_app = [sim_str,man_str,num2str(1),app_str,sta_str,'.txt']; man_2_fn_app = [sim_str,man_str,num2str(2),app_str,sta_str,'.txt']; man_3_fn_app = [sim_str,man_str,num2str(3),app_str,sta_str,'.txt']; man_4_fn_app = [sim_str,man_str,num2str(4),app_str,sta_str,'.txt']; man_5_fn_app = [sim_str,man_str,num2str(5),app_str,sta_str,'.txt']; dlmwrite(man_1_fn_app,man_1_app) dlmwrite(man_2_fn_app,man_2_app) dlmwrite(man_3_fn_app,man_3_app) dlmwrite(man_4_fn_app,man_4_app) dlmwrite(man_5_fn_app,man_5_app) man_1_fn_turn = [sim_str,man_str,num2str(1),turn_str,sta_str,'.txt']; man_2_fn_turn = [sim_str,man_str,num2str(2),turn_str,sta_str,'.txt']; man_3_fn_turn = [sim_str,man_str,num2str(3),turn_str,sta_str,'.txt']; man_4_fn_turn = [sim_str,man_str,num2str(4),turn_str,sta_str,'.txt']; man_5_fn_turn = [sim_str,man_str,num2str(5),turn_str,sta_str,'.txt']; dlmwrite(man_1_fn_turn,man_1_turn) dlmwrite(man_2_fn_turn,man_2_turn) dlmwrite(man_3_fn_turn,man_3_turn) dlmwrite(man_4_fn_turn,man_4_turn) dlmwrite(man_5_fn_turn,man_5_turn) man_1_fn_rec = [sim_str,man_str,num2str(1),rec_str,sta_str,'.txt']; man_2_fn_rec = [sim_str,man_str,num2str(2),rec_str,sta_str,'.txt']; man_3_fn_rec = [sim_str,man_str,num2str(3),rec_str,sta_str,'.txt']; man_4_fn_rec = [sim_str,man_str,num2str(4),rec_str,sta_str,'.txt']; man_5_fn_rec = [sim_str,man_str,num2str(5),rec_str,sta_str,'.txt']; dlmwrite(man_1_fn_rec,man_1_rec) dlmwrite(man_2_fn_rec,man_2_rec) dlmwrite(man_3_fn_rec,man_3_rec) dlmwrite(man_4_fn_rec,man_4_rec) dlmwrite(man_5_fn_rec,man_5_rec) status = 'Finished'

PAGE 130

118 APPENDIX E EXAMPLE OF SPSS OUTPUT FILE FO R MAXIMUM YAW DURING THE TURN PHASE OF MANEUVER 1 Within-Subjects Factors Measure: MEASURE_1 yawmaxa yawmaxb improv 1 2 Dependent Variable Between-Subjects Factors 39 32 0 1 agegp2 N Descriptive Statistics .3582633 .07513164 39 .3436297 .03921199 32 .3516679 .06163604 71 .4356249 .06452554 39 .4364078 .07731970 32 .4359777 .07005650 71 agegp2 0 1 Total 0 1 Total yawmaxa yawmaxb Mean Std. Deviation N Multivariate Testsc .566 90.058b 1.000 69.000 .000 .566 90.058 1.000 .434 90.058b 1.000 69.000 .000 .566 90.058 1.000 1.305 90.058b 1.000 69.000 .000 .566 90.058 1.000 1.305 90.058b 1.000 69.000 .000 .566 90.058 1.000 .011 .739b 1.000 69.000 .393 .011 .739 .136 .989 .739b 1.000 69.000 .393 .011 .739 .136 .011 .739b 1.000 69.000 .393 .011 .739 .136 .011 .739b 1.000 69.000 .393 .011 .739 .136 Pillai's Trace Wilks' Lambda Hotelling's Trace Roy's Largest Root Pillai's Trace Wilks' Lambda Hotelling's Trace Roy's Largest Root Effect improv improv agegp2 Value F Hypothesis df Error df Sig. Partial Eta Squared Noncent. Parameter Observed Powera Computed using alpha = .05 a. Exact statistic b. Design: Intercept+agegp2 Within Subjects Design: improv c.

PAGE 131

119 Mauchly's Test of Sphericityb Measure: MEASURE_1 1.000 .000 0 1.000 1.000 1.000 Within Subjects Effect improv Mauchly's W Approx. Chi-Square df Sig. Greenhous e-Geisser Huynh-Feldt Lower-bound Epsilona Tests the null hypothesis that the error covariance matrix of the orthonormalized transformed dependent variables is proportional to an identity matrix. May be used to adjust the degrees of freedom for the averaged tests of significance. Corrected tests are displayed in the Tests of Within-Subjects Effects table. a. Design: Intercept+agegp2 Within Subjects Design: improv b. Tests of Within-Subjects Effects Measure: MEASURE_1 .254 1 .254 90.058 .000 .566 90.058 1.000 .254 1.000 .254 90.058 .000 .566 90.058 1.000 .254 1.000 .254 90.058 .000 .566 90.058 1.000 .254 1.000 .254 90.058 .000 .566 90.058 1.000 .002 1 .002 .739 .393 .011 .739 .136 .002 1.000 .002 .739 .393 .011 .739 .136 .002 1.000 .002 .739 .393 .011 .739 .136 .002 1.000 .002 .739 .393 .011 .739 .136 .195 69 .003 .195 69.000 .003 .195 69.000 .003 .195 69.000 .003 Sphericity Assumed Greenhouse-Geisser Huynh-Feldt Lower-bound Sphericity Assumed Greenhouse-Geisser Huynh-Feldt Lower-bound Sphericity Assumed Greenhouse-Geisser Huynh-Feldt Lower-bound Source improv improv agegp2 Error(improv) Type III Sum of Squares df Mean Square F Sig. Partial Eta Squared Noncent. Parameter Observed Powera Computed using alpha = .05 a. Tests of Within-Subjects Contrasts Measure: MEASURE_1 .254 1 .254 90.058 .000 .566 90.058 1.000 .002 1 .002 .739 .393 .011 .739 .136 .195 69 .003 improv Linear Linear Linear Source improv improv agegp2 Error(improv) Type III Sum of Squares df Mean Square F Sig. Partial Eta Squared Noncent. Parameter Observed Powera Computed using alpha = .05 a. Tests of Between-Subjects Effects Measure: MEASURE_1 Transformed Variable: Average 21.772 1 21.772 3657.032 .000 .981 3657.032 1.000 .002 1 .002 .283 .596 .004 .283 .082 .411 69 .006 Source Intercept agegp2 Error Type III Sum of Squares df Mean Square F Sig. Partial Eta Squared Noncent. Parameter Observed Powera Computed using alpha = .05 a.

PAGE 132

120 LIST OF REFERENCES Boyce Thomas E., Geller Scott E. “An inst rumented vehicle assessment of problem behavior and driving style: Do younger males really take more risks?” Accident Analysis and Prevention 34 (2002): 51-64. Dingus, Thomas A., Hulse Melissa C. “Effects of age, system experience, and navigation technique on driving with an Advanced Traveler Information System” Human Factors: The Journal of the Huma n Factors and Ergonomics Society 39 (1997): 177-200. DOT HS 809 328. U.S. Department of Trans portation, National Highway Traffic Safety Administration. < http://wwwnrd.nhtsa.dot.gov/pdf/nrd30/ncsa/tsf2000/2000oldpop.pdf> (accessed June, 2004) Knodler Michael A., Noyce David A., Kacir Kent C., Brehmer Christopher L. “Evaluation of traffic signal displays for protected pe rmissive left-turn control using driving simulator technology.” Jo urnal of Transportation Engineering 131.4 (2005): 270-278. LabVIEW. Version 6.1. 2001. Natio nal Instruments. Austin, Texas. MATLAB. Version 6.1.0.450 Release 12.1. 1984-2001. The MathWorks, Inc. Natick, Massachusetts McGehee Daniel V., Lee John D., Rizzo Ma tthew, Bateman Kirk. “Examination of older driver steering adaptation on a high performance driving simulator.” International Driving Symposium on Human Factors in Driver Assessment, Training, and Vehicle Design (2001) (accessed February, 2006) Rizzo M., Jermeland J., Severson J. “Instr umented vehicles and driving simulators.” Gerontechnology 1.4 (2002): 291-296. Road & Track 53.6 (2002): 114-115. SPSS for Windows. Version 13.0. 1989-2004. SPSS, Inc. Chicago, Illinois. Uc E. Y., Rizzo M., Anderson S. W., Shi Q., Dawson J. D. “Driver route-following and safety errors in early Alzheimer’s disease.” Neurology 63 (2004): 832-837.

PAGE 133

121 Wierwille W. W., Gutmann F. “Comparison of primary and secondary task measures as a function of simulated vehicle dynamics an d driving conditions.” Human Factors: The Journal of the Human F actors and Ergonomics Society 20.2 (1978): 233-344. Wood J.M., Worringham C. “Quantitative assessment of driving performance in Parkinson’s disease” Journal of Ne urology, Neurosurgery, and Psychiatry 76 (2005): 176-180.

PAGE 134

122 BIOGRAPHICAL SKETCH Ethan Davis graduated from Pine View High School for The Gifted located in Osprey, Florida, in June 1999. He attend ed Florida Southern College in Lakeland, Florida, where he played 2 seasons of NC AA Div. II soccer. Etha n transferred to the University of Florida in Gainesville, Florid a where he graduated magna cum laude with a bachelorÂ’s degree in mechanical engineering in May, 2004. He will receive a Master of Engineering degree in May, 2006, from the University of Florida.