<%BANNER%>

Roundabout Modeling in CORSIM

Permanent Link: http://ufdc.ufl.edu/UFE0025029/00001

Material Information

Title: Roundabout Modeling in CORSIM
Physical Description: 1 online resource (82 p.)
Language: english
Creator: Elias, Aaron
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: corsim, modeling, roundabout, roundabouts, tsis
Civil and Coastal Engineering -- Dissertations, Academic -- UF
Genre: Civil Engineering thesis, M.E.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Roundabout construction in the United States is on the rise and traffic engineers need a good micro-simulation program to determine a roundabout?s impact on a transportation network. CORSIM is one of the most widely used micro-simulator in the United States, but its ability to accurately replicate roundabout operations is not well documented. This research compares field data from two study roundabouts to their simulation within CORSIM. This comparison showed that a roundabout can be modeled in CORSIM by making relatively few changes to CORSIM?s default values (such as gap acceptance), but the accuracy of the results varies by analysis period and roundabout approach. Based on the work completed in this research, recommendations for enhancing the software are made. Finally an interim guide to roundabout implementation within CORSIM is presented for use until the recommended changes to the software can be completed.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Aaron Elias.
Thesis: Thesis (M.E.)--University of Florida, 2009.
Local: Adviser: Elefteriadou, Ageliki L.
Local: Co-adviser: Washburn, Scott S.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0025029:00001

Permanent Link: http://ufdc.ufl.edu/UFE0025029/00001

Material Information

Title: Roundabout Modeling in CORSIM
Physical Description: 1 online resource (82 p.)
Language: english
Creator: Elias, Aaron
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: corsim, modeling, roundabout, roundabouts, tsis
Civil and Coastal Engineering -- Dissertations, Academic -- UF
Genre: Civil Engineering thesis, M.E.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Roundabout construction in the United States is on the rise and traffic engineers need a good micro-simulation program to determine a roundabout?s impact on a transportation network. CORSIM is one of the most widely used micro-simulator in the United States, but its ability to accurately replicate roundabout operations is not well documented. This research compares field data from two study roundabouts to their simulation within CORSIM. This comparison showed that a roundabout can be modeled in CORSIM by making relatively few changes to CORSIM?s default values (such as gap acceptance), but the accuracy of the results varies by analysis period and roundabout approach. Based on the work completed in this research, recommendations for enhancing the software are made. Finally an interim guide to roundabout implementation within CORSIM is presented for use until the recommended changes to the software can be completed.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Aaron Elias.
Thesis: Thesis (M.E.)--University of Florida, 2009.
Local: Adviser: Elefteriadou, Ageliki L.
Local: Co-adviser: Washburn, Scott S.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0025029:00001


This item has the following downloads:


Full Text

PAGE 1

1 ROUNDABOUT MODELING IN CORSIM By AARON ELIAS 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 FLORID A 2009

PAGE 2

2 2009 Aaron Elias

PAGE 3

3 To my friends and family for supporting me in my graduate endeavors

PAGE 4

4 ACKNOWLEDGMENTS I am indebted to a number of individuals for their help in completing this thesis. First I thank Lee Rodegerdts for his help in acquiring the data necessary to complete my research. I would also like to thank David Hale for all his help with my roundabout implementation in CORSIM. Thirdly, I thank Alexandra Kondyli for always being availabl e to answer questions, not only for this thesis, but throughout my time in graduate school Finally, I would like to thank my supervisory committee, Bill Sampson, Scott Washburn, and especially Lily Elefteriadou for providing guidance and support throughout my graduate studies

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................................... 4 LIST OF TABLES ................................................................................................................................ 7 LIST OF FIGURES .............................................................................................................................. 8 ABSTRACT ........................................................................................................................................ 1 0 CHAPTER 1 INTRODUCTION ....................................................................................................................... 11 Background .................................................................................................................................. 11 Problem Statement ...................................................................................................................... 11 Research Objectives .................................................................................................................... 12 Document Organization .............................................................................................................. 13 2 LITERATURE REVIEW ........................................................................................................... 14 Introduction ................................................................................................................................. 14 Modern Roundabouts .................................................................................................................. 14 Analysis Procedures .................................................................................................................... 18 United States Design Standards .......................................................................................... 18 HCM 2000 .................................................................................................................... 18 NCHRP Report 572 (Roundabouts in the United States) .......................................... 19 Linear Regression Based (Empirical) Method ................................................................... 22 Gap Acceptance -Based Australian Method ....................................................................... 24 Software Analysis Tools ............................................................................................................. 26 ARCADY and RODEL ....................................................................................................... 26 SIDRA .................................................................................................................................. 26 HCS+ .................................................................................................................................... 27 CORSIM ............................................................................................................................... 27 VISSIM ................................................................................................................................ 29 PARAMICS ......................................................................................................................... 32 Summary and Conclusions ......................................................................................................... 34 3 FIELD DATA .............................................................................................................................. 37 Introduction ................................................................................................................................. 37 Data Extraction Overview .......................................................................................................... 37 Three -Approach Roundabout: Port Orchard, Washington ....................................................... 40 Four -Approach Roundabout: Lothian, Maryland ...................................................................... 44

PAGE 6

6 4 CORSIM SIMULA TION ........................................................................................................... 48 Introduction ................................................................................................................................. 48 Network Implementation ............................................................................................................ 48 Inputting Roundabout Origin Destinations ............................................................................... 54 CORSIM Simulation Output Analysis ....................................................................................... 59 CORSIM Implementation Summary and Conclusions ............................................................. 62 5 RESEARCH CONCLUSIONS .................................................................................................. 67 Introduction ................................................................................................................................. 67 Recommended CORSIM Changes ............................................................................................. 67 User Friendliness and Animation Changes ........................................................................ 67 Traffic Simulation Changes ................................................................................................ 69 Summary of Recommendations .......................................................................................... 71 Recommendations for Future Research ..................................................................................... 72 APPENDIX: INTERIM GUIDE TO ROUNDABOUT MODELING IN CORSIM ................... 74 LIST OF REFERENCES ................................................................................................................... 80 BIOGRAPHICAL SKETCH ............................................................................................................. 82

PAGE 7

7 LIST OF TABLES Table page 2 1 LOS for roundabouts. ............................................................................................................. 22 3 1 Origin Destinations for the Port Orchard, WA site. ............................................................ 42 3 2 Entering and conflicting flow summary for the Port Orchard, WA site. ............................ 42 3 3 Summary of extracted field data from the Port Orchard, WA site. ..................................... 43 3 4 Origin Destinations for the Lothian, MD site. ..................................................................... 46 3 5 Entering and conflicting flow summary for the Lothian, MD site. ..................................... 46 3 6 Summary of extracted field data from the Lothian, MD site. ............................................. 47 4 1 Accepgap for right turning vehicles ...................................................................................... 54 4 2 Link volume comparison at the Port Orchard site. .............................................................. 57 4 3 Link volume comparison at the Lothian site. ....................................................................... 58 4 4 Comparison between C ORSIM and field data for the Port Orchard, WA site. ................. 61 4 5 Comparison between CORSIM and field data for the Lothian, MD site. .......................... 62 4 6 Average control delay values across the 8 time periods for the Port Orchard, WA site ... 65 4 7 Average control delay values across the 8 time periods for the Lothian, MD site. ........... 66

PAGE 8

8 LIST OF FIGURES Figure page 2 1 Modern roundabout design elements. ................................................................................... 15 2 2 Conflict point comparison. .................................................................................................... 16 2 3 Average delay for a roundabout compared to a signal. ....................................................... 17 2 4 An English roundabout with highly flared tangential entries. ............................................. 24 2 5 Example of a roundabout setup in CORSIM ........................................................................ 28 2 6 VISSIM curved link ............................................................................................................... 31 2 7 Example of priority rules bar locations ................................................................................. 32 2 8 A roundabout with four nodes modeled in PARAMICS ..................................................... 33 3 1 Example of z line used for control delay calculation .......................................................... 39 3 2 Overview of the Port Orchard, WA site.. ............................................................................. 41 3 3 Camera views of the Port Orchard, WA s ite. ....................................................................... 41 3 4 Overview of the Lothian, MD site ........................................................................................ 45 3 5 Camera Views of the Lothian, MD site. ............................................................................... 45 4 1 Four approach Lothian, MD arrival distributions. ............................................................... 49 4 2 Three approach Port Orchard, WA arrival distributions. .................................................... 49 4 3 Port Orchard, WA CORSIM network. .................................................................................. 50 4 4 Lothian, MD CORSIM network. ........................................................................................... 51 4 5 Port Orchard, WA CORSIM animat ion. ............................................................................... 52 4 6 Lothian, MD CORSIM animation. ........................................................................................ 53 4 7 CORSIMs logic for conditional turn movements at roundabouts ..................................... 56 A 1 Step 2 Adjust simulation duration ...................................................................................... 74 A 2 Step 3 Change arrival distribution ..................................................................................... 75 A 3 Steps 4 7 Construction of network nodes .......................................................................... 76

PAGE 9

9 A 4 Step 8 Internal link modifications ...................................................................................... 76 A 5 Step 10 Generation of traffic at the 8000 nodes ................................................................ 77 A 6 Step 11 External approach turn movement coding ........................................................... 77 A 7 Step 12 Internal approach turn movem ent coding ............................................................ 78 A 8 Step 13 Internal approach conditional turn movement coding ........................................ 78 A 9 Step 14 Modifications to gap acceptance .......................................................................... 79

PAGE 10

10 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 ROUNDABOUT MODELING IN CORSIM By Aaron Elias August 2009 Chair: Lily Elefteriadou Co C hair: Scott Washburn Major: Civil Engineering Roundabout construction in the United States is on the rise and traf fic engineers need a good micro -simulation program to determine a roundabout s impact on a t ransportation network. CORSIM is one of the most widely used micro -simulator in the United States but its ability to accurately replicate roundabout operations is not well documented This research compares field data from two study roundabouts to their simulation within CORSIM. This comparison show ed that a roundabout can be modeled in CORSIM by making relatively few changes to CORSIMs default values (such as gap acceptance) but the accuracy of the results varies by analysis period and roundabout app roach. Based on the work completed in this research, recommendations for enhancing the software are made. Finally an interim guide to roundabout implementation within CORSIM is presented for use until the recommended changes to the software can be comple ted.

PAGE 11

11 CHAPTER 1 INTRODUCTION Background The adoption of a mandatory give -way rule at circular intersections by the United Kingdom in 1966 was the beginning of the modern roundabout (Robinson et al 2000). Roundabouts can improve safety and decrease delay, w hich result ed in widespread implementation of this intersection type throughout Europe and Australia. Today there are in excess of 35,000 modern roundabouts in the world ( Jacquemart 1998). The success of the modern roundabout in Europe and Australia has led to a rise in their construction and operation within the United States. It was not until the early 1990s that American transportation engineers began looking at the modern roundabout as a viable alternative to other forms of intersections. H esitation of American engineers to construct modern roundabouts is explained in part by the poor performance of circular intersections constructed in the first half of the 20th century These circular intersections did not have the same design standards as the modern roundabout and are today known as non conforming traffic circles. Since the first two modern roundabouts were constructed in Summerlin, Nevada in March of 1990 ( Jacquemart 1998), there has been significant growth in the construction and design of roundabouts and there are in excess of 300 modern roundabouts in the United States today ( Rodegerdts et al 2007) Problem Statement The rapid growth of roundabouts in the United States has outpaced the development of analytical tools and software used to analyze their operational characteristics such as capacity and delay. This forced early American roundabout practitioners to depend on analysis procedures developed in other countries and while these methods did provide some level of

PAGE 12

12 analysis, they were not designed for United States driving conditions or drivers. The new Highway Capacity Manual will alleviate many of these problems by providing equations of capacity and delay based on the research of NCHRP (National Cooperative Highway Research Program) pr oject 3 65 Roundabouts in the United States The focus now needs to be on ensuring accurate reproduction of a roundabout in micro -simulation software. CORSIM is an example of this type of software and is the most widely used micro -simulator in the Unit ed States. CORSIM s widespread use as a traffic simulation software is due to its original development by the Federal Highway Administration (FHWA) and the resulting relatively low cost. However, CORSIMs ability to accurately model roundabouts is unknown since little research has been completed to verify the results of a simulated roundabout The ever increasing numbers of roundabouts along with the widespread use of CORSIM s uggests this software package needs to be evaluated on its ability to accurately replicate roundabouts Research Objectives The objective of this research is to evaluate CORSIMs ability to model roundabouts and based on this evaluation, recommend changes to improve its modeling ability The implementation of these recommendations wi ll provide transportation engineers the tools they need to analyze the impact of roundabouts on network performance. The tasks that were conducted to achieve this objective were: Completed a literature review on current analysis techniques and software pac kages used to analyze roundabouts. Collect ed field data from two roundabouts in the United States operating under a variety of different flow conditions. Simulat ed these roundabouts in CORSIM.

PAGE 13

13 Perform ed a macroscopic comparison between CORSIMs output a nd the observed performance measures from the field. M a de recommendations based on the findings of the previous tasks, on how to improve CORSIMs roundabout modeling ability. Crea ted an interim guide to roundabout implementation using the current versio n of CORSIM Document Organization Chapter 2 presents a literature review on modern roundabouts, analysis procedures used to get performance measures of roundabouts, and a review of roundabout software analytical tools. Chapter 3 describes the collection of field data from two roundabout sites, a three approach roundabout and a four approach roundabout Chapter 4 presents the implementation of the study roundabouts into CORSIM and compares the results between the field data collection and CORSIMs output. The final chapter presents the recommended enhancements to CORSIM based on this research and provides an interim guide to CORSIM roundabout modeling using the current software version.

PAGE 14

14 CHAPTER 2 LITERATURE REVIEW Introduction This chapter reviews the usage of th e modern roundabout in the United States including sections on analytical analysis methodologies and an overview of the software used to study their operational parameters. Modern Roundabout s The use of circulatory intersection design s has been around sinc e the construction of Columbus Circle in New York City, which opened in 1905. With the implementation of this design, many other cities built large traffic circles or rotaries. Many problems were discovered with this new style of intersection after their construction and use These initial designs gave right -of -way to the entering traffic resulting in high speed entries and higher crash volume experiences. In addition to increased crashes, these circles would lock up at higher traffic volumes creating c ongestion. The high crash experience and congestion in the circles led to rotary designs falling out of favor in America after the mid 1950s (Robinson et. al., 2000). These types of traffic circles are known today as nonconforming traffic circles and in corporate one or more of the following operational or design elements, which make them a less than ideal intersection type ( Jacquemart 1998): Entering traffic had the right -of -way At higher volumes this locks up the circle. Entries were tangential to t he circle This encourages high entering speeds and reduces the safety benefits. Pedestrians crossed onto the central island This is unsafe for pedestrians and disruptive for drivers. The through road cut through the circle Capacity, fluidity, and s afety benefits are lost by the need to signalize the central intersection.

PAGE 15

15 Circulating traffic was controlled by a traffic signal or stop sign This decreases the fluidity of circulating traffic and can lock up the circle. Parking was permitted in the circle This reduces the capacity and safety of the circle by adding friction and conflicts. With these nonconforming traffic circles facing so many problems, many engineers began to think of ways to improve the circulatory intersection design. The answ er came in the 1960s with the advent of the modern roundabout in the United Kingdom. The most important change the modern roundabout incorporated into circulatory intersections is the give -way rule which requires all entering vehicles to yield to the cir culating traffic within the roundabout. This new rule prevents the circle from lockingup. With the new rule in place, roundabout geometric design began to change to the modern roundabout as seen in Figure 2 1. Figure 2 1 Modern roundabout d esign e lements [Adapted from Robinson et al 2000. Roundabouts: an informational guide Report FHWA RD 00 067 (Page 6, Exhibit 1 2). Federal Highway Administration, Washington DC. ] Nonconforming traffic circles lik e Columbus circle were designed to maximize the major street movements and the weaving distances to keep traffic moving. As a consequence, many entries were tangential to the circulating traffic leading to high speed entry and a higher collision

PAGE 16

16 risk. As the modern roundabout no longer uses weaving to move traffic through the circle, approaches were redesigned to incorporate two key elements. The first is aligning the centerline of every approach with the center of the central island or as close as possi ble to center. The second is the installation of splitter islands which deflect traffic to the right around the central island. These two features force slower entry speeds and increased roundabout safety. In addition to these two features, the main safe ty benefit of roundabouts is the reduction in the number of confli ct points. A conventional four approach intersection has thirty two points of conflict while a modern roundabout has only eight. A comparison of these conflict points is shown in Figure 2 2 Not only are there fewer conflict points, but none of the conflict points are crossing conflicts where turning movements cross opposing traffic streams. These are the most severe conflicts leading to high speed, right angle collisions at other inter section types. Figure 2 2 Conflict p oint c omparison [Adapted from Robinson et al 2000. Roundabouts: an informational guide Report FHWA RD 00 067 (Page 106, Exhibit 5 2). Federal Highway Administrati on, Washington DC. ] The slower speeds and fewer conflict points at a modern roundabout can have a significant impact on the safety of an intersection. One of the most comprehensive roundabout safety studies was conducted on 12,000 roundabouts in France. It was reported that signalized intersections had accident frequencies four times higher than roundabouts carrying similar traffic

PAGE 17

17 flow s (Guichet, 1997). A second study by the Insurance Institute for Highway Safety found there were large reductions in th e number of injury crashes, especially those involving incapacitating or fatal injuries (Persaud et al., 2000). Roundabouts have been reported to have tremendous advantages in capacity and delay reductions over other intersection types and they can handle very high traffic flows. It is not uncommon for roundabouts in the United Kingdom to carry in excess of 6,000 vehicles per hour (Todd, 1991). Due to its ability for high capacity, it is possible to replace signals with roundabouts thereby reducing inters ection delay. A Federal Highway Administration report on roundabouts found that a typical roundabout saved drivers 15,000 vehicle hours per year over a signal with 33% of vehicles making a left turn. (Robinson et al 2000) A graphical representation of the delay savings can be seen in Figure 2 3 which shows average delay of a roundabout compared to a signal. Roundabouts are especially good in delay savings during the off-peak period as the vehicles do not have to wait for a signal to cycle through. F igure 2 3 Average d elay for a r oundabout c ompared to a s ignal [Adapted from Robinson et al 2000. Roundabouts: an informational guide Report FHWA -RD 00067 (Page 63, Exhibit 3 7). Federal Highway Administ ration, Washington DC. ]

PAGE 18

18 Analysis Procedures Although roundabouts have been used for many years in countries throughout the world, they are still relatively new to the United States. For this reason, there is no uniform method of performance analysis usabl e by age ncies considering roundabouts. This section presents the current and proposed analyses procedures, analyzes the two prominent methods (the linear regression based United Kingdom method and the gapacceptance based Australian method), and explores the current leading software for roundabout analysis. United States Design Standards The most common operational analysis guide in the United States is the Highway Capacity Manual (HCM). The current edition was published in 2000 (HCM 2000) and incorporate d for the first time a small section on roundabouts with the intent to begin a standardization of roundabout analysis techniques in the United States. This standardization is needed because many practicing engineers use roundabout design criteria from oth er countries which does not always correspond to driving conditions in the U.S. A more substantial guide to roundabouts will be incorporated, based on the National Cooperative Highway Research Programs (NCHRP) project 3 65 and its summary report 572, wi th the new HCM in 2010 ( 2010 HCM). Looking at the current and future analysis procedures in the HCM will be the focus of the next section. HCM 2000 Found in Chapter 17 Part C, the roundabout section of HCM 2000 has four pages and one example problem. Uni ted States practitioners are warned that due to limited experience and little field data, the roundabout analysis procedures described should be used as basic guidelines until more research can be completed. This lack of field data has led the HCM to use a gap acceptance based model rather than the empirically derived linear regression models.

PAGE 19

19 With roundabouts requiring all drivers to make right turns, the HCM assumes the gap acceptance characteristics of drivers for a two way stop control (TWSC) intersect ion are similar to the gap acceptance at a single lane roundabout (HCM 2000). HCM 2000 gives the following formula for capacity of single lane roundabouts: 3600 / 3600 /1f c c ct v t v c ae e v c (2 1) where, ca = approach capacity (veh/h) vc = conflicting circulating tr affic (veh/h) tc = critical gap (s), and tf = follow up time (s) HCM 2000 concludes that the critical gap at roundabouts is bounded with 4.1 seconds being the upper bound and 4.6 seconds being the lower bound. Follow up time is also bounded with 2.6 seconds being the upper and 3.1 seconds being the lower bound. An approach capacity can be determined given follow up time, critical gap, and the volume of circulating vehicles. While HCM 2000 does make an attempt at helping practitioners derive a method for a nalyzing roundabout capacity, it is greatly lacking in versatility and field data. It does not account for multilane roundabouts, provide any means for determining control delay and queues, or account for geometric design such as number of entry lanes whi ch has a clear effect on the capacity of a roundabout entry (Rodegerdts et al 2007). NCHRP Report 572 addresses many of these issues. NCHRP Report 572 (Roundabouts in the United States) NCHRP 572 summarizes the findings of NCHRP project 3 65 where the a pplication of roundabouts in the United States was studied. This project developed a rough draft for a new

PAGE 20

20 chapter in the 2010 HCM. The main findings of this report dealing with the operational nature of roundabouts were: revised equations of capacity, a new control delay equation, and LOS criteria for each approach. There were two main changes to roundabout capacity analyses over the method currently used by HCM 2000. The first is that field data was used to determine capacity equations resulting in a b etter calibrated model. Second, HCM 2000 did not have an equation for a multilane roundabout which was rectified in NCHRP 572. Other findings of NCHRP 572 show that United States drivers use roundabouts less efficiently than the predicted outputs from ot her analysis methods from around the world (Rodegerdts et al 2007). This is most likely caused by the newness of roundabouts to the United States and the learning curve of drivers in navigating roundabouts. The general form of the capacity equation for single lane roundabouts is: ) 0010 0 exp( 1130cv c (2 2) where, c = entry capacity (passenger car units (pcu/h) c = conflicting flow (pcu/h) To make allowances for driver behavior in different areas, the authors of NCHRP 572 developed a capacity calibra tion model in the form of: cv B A c exp (2 3) where, c = entry capacity (pcu/h) A = 3600 / tf B = ( tc f / 2 ) / 3600

PAGE 21

21 c = conflicting flow (pcu/h) tf = follow up headway (s) tc = critical headway (s) In addition to these single lane rounda bout capacity equations, a two lane roundabout empirical equation was done for the critical lane taking the form: c critv c 0007 0 exp 1130 (2 4) where, ccrit = entry capacity of critical lane (pcu/h) c = conflicting flow (pcu/h) Another major improvement t o roundabout performance analyses over the HCM 2000 method is an equation for control delay. With control delay known, LOS can be determined for an intersection allowing current transportation practitioners to make a more informed choice when selecting an intersection type. This control delay model and the corresponding LOS thresholds developed by the NCHRP 572 team are presented in Table 2 1 and E quat ion 2 5 : T c v c c v c v T c d 450 3600 1 1 900 36002 (2 5) where, d = average contr ol delay (s/veh) c = capacity of subject lane (veh/h) t = time period (h: T = 1 for 1 -h analysis, T = 0.25 for 15-min analysis)

PAGE 22

22 Table 2 1 LOS t able for r oundabouts. Level of Service Average Control Delay (s/veh) A 0 10 B > 10 15 C > 15 25 D > 25 35 E > 35 50 F > 50 NCHRP 3 65 adds a substantial amount of available information to the transportation professional looking to implement a roundabout in the United S tates than was previously available using the basic equations provided in HCM 2000. These improvements greatly enhance the roundabout analysis guidelines for the United States allowing for accurate analytical analysis of roundabouts based on the latest st udy Linear Regression Based (Empirical) Method The lack of a uniform roundabout analysis guideline in the United States led early roundabout implementers to use a variety of methods in determining capacity estimates. The two leading methods used by pract itioners are the gap acceptance and linear regression methods. Used in the United Kingdom, the linear regression based empirical method is statistically derived and based on a large number of capacity measurements at saturated roundabouts ( Jacquemart 1998). It takes the form: (Kimber, 1980): c c eQ f F k q max (2 6) w here, qe,max = maximum entry flow (veh/h) Qc = circulating flow (veh/h) 2 k = 1

PAGE 23

23 2 S = (e e = entry radius ( m) -width (m) l = effective flare length (m) r = entry radius (m) = entry angle (degrees) S = measure of the degree of the flaring D = inscribed circle diameter (m) The United Kingdom empirical method incorporates far more geometric par ameters than the typical inscribed diameter, entry width, and circulating width used by most other capacity estimate methods. However, this regression based empirical method is not without problems. The model developed in the UK comes from a large sample of roundabouts at capacity. As a consequence, the UK linear regression model underestimates capacity for low circulating flows and overestimates for high circulating flows (Akelik, 2003). Another issue with using the UK models in the United States is th at most roundabouts in the UK are designed differently than the typical practices found in Australia and the United States. Highly flared roundabout entries leading to tangential entries in the UK are more likely to cause merging of entering traffic rathe r than yielding to circulating traffic ( Figure 2 4 ).

PAGE 24

24 Figure 2 4 An English r oundabout with h ighly flared t angential e ntries. Gap Acceptance-Based Australia n Method Australia uses a different method than the empirical method employed in United Kingdom, which is based on gap acceptance The theory behind this capacity method is to identify minimum headways between two vehicles in which a vehicle enter s into the circulating roundabout flow from an external approach. This critical gap acceptance is dependent on follow up time, circulating flow, number of circulating lanes, and the average circulating lane width ( Rodegerdts et al 2007). The most recent capacity expressions available come from Akelik et. al. (Akelik et. al., 1999) is : m g od eq q f q maxmax (2 7)

PAGE 25

25 wher e, c c c c gq q q exp 3600 5 0 3600 1 3600 m e mn q q 60 min cd qd qc odp p f f 1 (veh/h) lane entry an for flow entry maximummax ,eq (veh/hr) flow entry minimum gq (veh/h) flow g conflictin cq (veh/h) flow arrival entry eq factor adjustment d o odf used) (0.6 0.8 to 0.5 qd cdp p (veh/min) flow entry minimum mn flow g conflictin in lanes of number cn 2 for 2 1 1 for 0 2 (s) traffic g circulatin in headway minimum c c cn n else 49 / 98 0 3600 / 3600 1 3600 (veh/s) factor on distributi headway arrival c c c c c c c cq for q q 2 for 3600 / 0 3 exp 1 for 3600 / 0 5 exp vehicles g conflictin unbunched of proportion c c c c cn q n q

PAGE 26

26 (s) headway up follow In order to maintain satisfactory traffic flow through a roundabout, the Australian roundabout design guide (AUSTROADS 1993) recommends a degree of saturation less than 0.8 to 0.9. Software Analysis Tools A number of different software packages can be used to analyze operational parameters of roundabouts. These software packages can be broken down into two distinct categories: 1) the deterministic packages (ARCADY, RODE L, SIDRA, and HCS+) and 2) the simulation packages (CORSIM, VISSIM, and PARAMICS). The following sections explore these software programs. ARCADY an d RODEL Based on the United Kingdoms linear regression formula ARCADY (Assessment of Roundabout CApacity a nd DelaY) was developed by the Transportation Research Lab (TRL). Currently on release 6, ARCADY had been in use over the past 20 years and is used to predict capacity, queue length, delays, and crash frequencies as a function of geometry ( Jacquemart 1998). Another popular UK software package is RODEL RODEL (ROundabout DELay) was first developed in 1987 and provides flexibility for experimenting with the geometric design of roundabouts. RODEL outputs capacity estimates, average and maximum delay, queues for each approach and an estimate of overall delay. The biggest difference between RODEL and ARCADY is that RODEL allows the user to select a confidence interval for capacity meeting or exceeding the desired value. If the confidence level is set to 50% both programs will return the same result ( Jacquemart 1998). SIDRA A survey conducted for NCHRP 264 found the SIDRA software to be the most commonly used roundabout analysis software in the United States with 46% of respondents

PAGE 27

27 choosing it ( Jacquemart 1 998). SIDRA was developed by ARRB Transport Research Limited. Since this is an Australian -based organization, this software uses gap acceptance techniques as described previously in the determination of a roundabouts capacity. It is important to note that while ARRB provided the fundamentals of roundabout analysis to the SIDRA software package, this software has since been modified to account for overestimation of capacity by use of simulation and professional judgment by Akelik (Rodegerdts et al 2007). HCS+ Highway Capacity Software or HCS+ is a deterministic package used for operational and planning applications, and it includes features to calculate roundabout capacity. Since it uses HCM 2000 procedures it can be considered the official analysis method of United States practitioners. (Kinzel and Trueblood, 2004) Using the parameters of critical gap and follow up time along with turning movements, it computes the capacity of each approach. Since the HCM 2000 does not define equations for delay it is not an output of HCS nor is the expected queue length as this is a function of delay. Another issue with this analysis software is that it makes no provisions for roundabouts with more than four approaches or more than two circulating lanes. CORSIM CORridor SIMulation or CORSIM is a combination off two microscopic simulation models, NETSIM and FRESIM. NETSIM simulates traffic on an urban street while FRESIM is a simulator for freeway operations Originally developed by the Federal Highway Administ ration (FHWA) it is now maintained and distributed by the McTrans at the University of Florida. Courage states: This software product represents the results of a substantial research effort and has gained widespread use in the USA for traffic control sys tem analysis. (Courage, 1997) Even though it has gained widespread usage in the United States, CORSIM does have a number of limitations and one of these is the simulation of roundabouts.

PAGE 28

28 Roundabouts are not explicitly modeled in CORSIM but can be indirec tly modeled using current software elements. CORSIM uses a link and node structure to model a transportation network with nodes being intersections and links representing the connecting roadways. At its most basic level, a roundabout is a series of yield controlled T intersections connected by a circular roadway. To model this in CORSIM, the user must establish a node for each approach then connect them by a one -way link segment in a counterclockwise direction. An examp le of a four approach roundabout can be seen in Figure 2 5 In implementing a roundabout in CORSIM, Joseph indicated credible looking roundabout performance measures could be obtained from CORSIM as long as traffic volumes are reasonably well balanced. (Joseph, 1996) The FHWA does not recommend this technique if the purpose of the analysis is directly affected by the roundabout operation. However, if an adjacent intersection to a study corridor is a roundabout, it may be possible to model that intersection as a roundabout without jeopa rdizing the main objectives of the analysis. (Holm et.al., 2007) Figure 2 5 Example of a r oundabout s etup in CORSIM.

PAGE 29

29 While it is possible to get credible looking roundabout performance measurers from CORSIM, the softwares output has not been compared to an actual roundabout. In addition to this, a number of other limitations in roundabout modeling in CORSIM have been discovered (Holm et.al., 2007): The minimum link length in CORSIM is 50ft which inc ludes the upstream intersection. This can cause a significant problem when it comes to modeling very small roundabouts. Short lengths are also difficult to curve into realistic looking roundabouts. Courage (1997) adds the following limitations: Approach g eometrics to a roundabout may be unique to each approach; however some parameters such as critical gap, follow up time, etc. are not approach -specific in CORSIM. An actual roundabout user may yield unnecessarily to a circulating vehicle that is about to e xit due to the uncertainty in the intentions of the circulating vehicle. CORSIM does not replicate this real -world situation. Short links make the gap acceptance process difficult to simulate. Current coding of roundabouts in CORSIM is both tedious and s ubject to errors given the manual nature of implementing the link node structure necessary for roundabouts. CORSIM does not provide the option for specifying roundabout parameters like critical gap and follow up time that is reported in the literature. W hile CORSIM is one of the most popular micro-simulators it has some problems. This is especially true for the simulation of roundabouts since they can only be modeled in an indirect way. This may or may not give the user reliable data. Given its widesp read use by practicing engineers and the increasing popularity of roundabouts in the United States, CORSIM needs to be upgraded to specifically handle roundabouts. VISSIM VISSIM is a micro -simulation program developed by the University of Karlsruhe in Germany during the early 1970s. Today it is maintained and distributed by PTV America

PAGE 30

30 (formerly Innovative Transportation Concepts.) Since it is able to model many different modes (i.e. buses, light rail, heavy rail, trucks, pedestrians, bicyclists, etc.) al ong with advanced transportation controls like ramp metering, transit signal priority, and dynamic lane control signals, VISSIM is one of the most advanced micro -simulators on the market. The flexibility and advanced nature of VISSIM make it one of the ma in micro -simulators used when modeling a roundabout. Trueblood (Trueblood, 2003) identified four main features of VISSIM that allow it to model roundabouts realistically: Link and Connectors Routing Decisions Priority Rules Reduced Speed Zones Rather than the traditional link and node feature used by CORSIM and other microsimulator programs, VISSIM uses a link and connector system. The link and node type of simulator puts the main emphasis on the node (intersection) with the links being built between two nodes. VISSIM, in contrast to a program like CORSIM, puts the emphasis on the links with the connectors connecting the different links. Basically connectors can be thought of as ramps onto links. (Trueblood, 2003) This is beneficial to the modeli ng of roundabouts because a VISSIM link can have several internal inflection points ( Figure 2 -6 ) without affecting the simulation of traffic flow. This not only leads to more realistic looking roundabouts, but negates the short link length problem in CORS IM that may cause car -following logic difficulties.

PAGE 31

31 Figure 2 6 VISSIM curved l ink [Adapted from Trueblood M. and Dale J., 2003. Simulating roundabouts with VISSIM. In: Proceedings of the 2nd Urban Stree t Symposium Anaheim (Page 5, Figure 2)] The second main feature allowing VISSIM to model roundabouts is its ability to control routing decisions. VISSIM allows users not only to code a specific path through a roundabout, but also codes the volume on eac h route as either a percentage of the total volume or the actual volume. CORSIM only allows users to specify a percentage of traffic continuing in the roundabout or leaving the roundabout at each node individually. VISSIM allows the user to specify these ODs and also, which lane a vehicle uses to complete its OD on multilane roundabouts. The third feature of VISSIM is the use of priority rules. Priority rules are represented by at least two bars with a red one placed at the interrupted movement and one or more green bars placed at the interrupting movement ( Figure 2 7 ). This configuration allows modelers to specify where a vehicle will yield (red bar) and the conflict location that a vehicle will use to judge critical gap distance (green bar). Modelers are also able to adjust gap acceptance times depending on the vehicle type so that a large truck will require a larger gap than a passenger car.

PAGE 32

32 Figure 2 7 Example of p riority rules b ar l ocations [Adapt ed from Trueblood M. and Dale J., 2003. Simulating roundabouts with VISSIM. In: Proceedings of the 2nd Urban Street Symposium Anaheim (Page 8, Figure 8)] The final feature of roundabout coding within VISSIM is the reduced speed zones. Since circulating traffic within the roundabout is expected to be between 1525 miles per hour depending on the size, it is necessary to code these speed decreases within the roundabout for realistic performance measures. This is accomplished by inputting speed zones on the circular roadway of a roundabout. VISSIMs greatest strength in modeling roundabouts is its flexibility. While its link connector network structure makes it more time consuming to construct, it allows for fine tuning the gap acceptance parameters for ea ch approach. (Stanek and Milam, 2005) With great flexibility, features that enable accurate modeling of roundabouts, and the only foreseeable downside being a lengthy coding process, VISSIM can be said to be one of the best microsimulators for roundabout simulation. PARAMICS PARAllel MICroscopic Simulation or PARAMICS is distributed by Quadstone from the United Kingdom. It was developed as part of large research and development projects under the

PAGE 33

33 European Community I project. (Oketch et.al., 2004) It is composed of six components with the modeler and analyzer being the core two needed to run the program. The others (processor, programmer, monitor, and estimator) can be included depending on the version ordered. It is similar to CORSIM in that it has a link node structure. PARAMICS uses a series of sub -nodes to replicate each approach to the roundabout. An example of this can be seen in Figure 2 8 where the roundabout is node 22 with sub-nodes a through d indicating each approach. Lanes can also be coded within PARAMICS roundabout lane editor to more closely model the lane markings and turning lanes found at the approaches and circulating links. Visibility is another special feature that can be coded at each roundabout sub -node within PARAMICS. This allows vehicles to accept gaps within the circulating flow depending on what the vehicles driver can see. Finally, PARAMICS is able to model short flared approaches, impact of geometry on speed within the roundabout, and uneven use of approach lane s due to turning proportions where other software may have problems. Figure 2 8 A r oundabout with four n odes m odeled in PARAMICS. [Adapted from Oketch T., Delsey M., and Robertson, D. 2004. Evaluation of performance of modern roundabouts using PARAMICS micro-simulation model. In: Proceedings of the TAC Annual Conference, Quebec City (Page 5, Figure 1)]

PAGE 34

34 The one major advantage PARAMICS has over CORSIM is its ability to model roundabouts directly with its series of sub -nodes depicting each approach. Oketch et.al concluded that PARAMICS offers a viable method for analyzing roundabouts. However, they do stress that further analysis is required along with validation of their model results using existing r oundabouts. (Oketch et.al., 2004) Summary and Conclusions NCHRP 3 65 is the most comprehensive report on roundabouts in the United States todate. The capacity equations outlined by the report were developed specifically for U.S. roundabouts and drivers b ased on video recordings of 31 sites within the U.S. These new capacity and delay equations will become the basis for all future operational analysis of roundabouts in the United States. There are two distinct software package types for analyzing the oper ations of a roundabout. ARCADY, RODEL, SIDRA, and HCS+ are examples of the deterministic type. ARCADY and RODEL provide delay, capacity and queue estimates, but their equations are based on British driver behavior and British roundabout designs which are different than those of U.S. roundabouts. SIDRA, the leading deterministic software package in the U.S., is based on the Australian gap acceptance method. SIDRA also outputs delay, capacity, and queue estimates and though there has been little field eva luation, experience has shown that its results fit U.S. single lane roundabouts. (Flannery et.al., 1998) However, roundabouts can have more than a single lane and SIDRA is still based on Australian roundabout experience. The final deterministic software is HCS+ which is still using the outdated methods from the 2000 HCM for roundabout analysis. The publication of the 2010 HCM will allow transportation professionals to deterministically analyze a roundabout using methods developed through research conduc ted in

PAGE 35

3 5 the United States. This will result in more accurate deterministic analys i s of roundabout operations. However, simulation is just as vital to roundabout analysi s as deterministic methods. Can a roundabout be implemented at this location effective ly and how will it affect the network as a whole are both questions that simulation can help to answer. Though there are a number of these programs on the market each has its own strengths and weaknesses. VISSIM is perhaps the most capable simulator progr am for modeling roundabouts. This is a result of its flexibility and its link -connector structure rather than the linknode structure found in other programs. The link -connector system allows for a more realistic looking network because the links are the controlling input allowing a coder to make the links more circular. The other two advantages of VISSIM include the ability to code origin -destination movements through the roundabout and the ability to set up priority rules allowing for more control of gap acceptance characteristics at each approach. The biggest disadvantage s of VISSIM are that it takes a long time to code all the special features needed to correctly model roundabout operations and it is an expensive program PARAMICS is similar to CORSIM but when it comes to modeling roundabouts it has an advantage in that it can directly model them through the use of sub-nodes. Using roundabout specific defaults which are automatically implemented PARAMICS is the easiest of the three software packages reviewed to implement roundabouts. However, it has not gained the widespread use of CORSIM and it i s not as detailed as VISSIM. CORSIM is the final simulation software reviewed and is one of the most popular in use in the U.S. This popularity stems from its relatively low cost and it s initial development by the FHWA. With many transportation professionals using this software there is great need for the software to be modified to easily model roundabouts and provide realistic performance

PAGE 36

36 measures. Wheth er these modifications can be accomplished through a few minor corrections such as easier implementat io n of origin -destinations or something more substantial like modifications to its car -following logic is something that needs to be investigated.

PAGE 37

37 CHAPTER 3 FIELD DATA Introduction F ield data are needed for comparison between CORSIMs output and the actual operations of the roundabout being simulated to determine CORSIMs ability to model roundabouts correctly. F ield data for this research were extracted from NCHRP 3 65s video recordings of two roundabout sites located in Port Orchard, Washington and Lothian, Maryland. A description of the data extracted, techniques of extraction, and site overviews are presented in this chapter. Data Extraction Overview A total of two hours of data for each roundabout site w ere extracted for the comparison between CORSIMs simulation and actual roundabout operations. These data can be grouped into two categories: 1) data required to simulate a roundabout in CORSIM and 2) data re quired to compare CORSIMs output to the field data. Data required for the CORSIM simulations are : Approach volumes for each leg of the analyzed roundabouts Origin -destinations of all traffic Geometric information including diameter and distance between ea ch approach on the circular roadway Speed information for each approach along with that of the circulating traffic Data required for comparisons with CORSIMs output are : Approach delay Average queue Maximum queue

PAGE 38

38 Approach volumes and origin -destinations for each leg of the roundabouts were collected using video recordings from an omnidirectional camera that observed the entire roundabout. This was accomplished using the traffic counting software DAITA (University of Florida, 2007) To simplify the extra ction of origin -destinations, the number of vehicles making a u-turn at the roundabouts was assumed to be negligible and counted as left turning movements. Collection of origin destinations allowed for the computation of entering and conflicting flow at e ach leg of the two roundabouts. This information can then be input into CORSIM for replication of traffic on each approach. Geometric characteristics including d iameter of the circulating roadway and the link lengths between each approach to the roundabou t were measured using Google Earth satellite images and the ruler measuring tool provided in the program. These values were used to properly size the roundabout within the CORSIM network all owing it to accurately simulate both the straight line and curved distance between roundabout nodes. The final piece s of data needed to implement the se roundabouts in CORSIM are the speeds of the vehicles using the network. Using Google Earth images the speed limit signs on the approaches to the roundabouts were obtai ned and input into CORSIM. The speed within the circulating roadway was determined first by establishing the distance a left turning vehicle traveled between the entry yield line and the exiting crosswalk using Google Earth satellite images. This number was then divided by t he number of seconds a vehicle needed to cover this distance. This was done for a sampling of thirty vehicles and the average of these was the speed of the circulating roadway. Approach delay is the main comparison used to test CORSI Ms ability to model roundabouts accurately. This research used a similar method of collecting control delay for each

PAGE 39

39 approach as used by NCHRP 3 65 researchers. Two reference lines were used with a Z line positioned upstream of a roundabout entry and th e yield line to mark a roundabout entry event as shown in Figure 3 1 Each vehicle was timed from when it crossed the upstream Z -line to entry into the roundabout by crossing the yield line. This time in system was collected using a program called Traf fic Tracker. A sampling of times for thirty unopposed vehicles was used to get the average free flow time due to geometric considerations between these two lines. Delay time for each vehicle was then calculated as their time in system minus the free flow time. Figure 3 1 Example of z -line u sed for c ontrol d elay c alculation The last of the data extracted from the video recordings for comparison were the average and maximum queue for the roundabout a pproaches using data from each vehicle including their approach delay, upstream Z -line entry time, and roundabout entry time. First, all vehicles not experiencing delay were eliminated from the data set because these vehicles cannot be part of a queue. T he second step was to filter each of the remaining vehicle Z -entry times into fifteen second bins covering the entire two hours of data collection. The same was done for their entry Z

PAGE 40

40 into the roundabout. A cumulative tally of total vehicles entered and to tal vehicles departed for every fifteen seconds was then calculated based on their entry and exit times. Finally, the total number of vehicles departed was subtracted from the total number of vehicles entered to get the standing queue at the end of each f ifteen second interval The average and maximum value of these queued vehicles was then taken and used in the comparison with CORSIMs output for average and maximum queue during the two hours of collected data Three -Approach Roundabout: Port Orchard, Wa shington The first site, a three approach roundabout in Port Orchard, Washington, was selected because data collection included all three approaches and an omnidirectional camera to observe the circulating traffic. Figure 3 2 shows a satellite image of th is site and Figure 3 3 shows the camera coverage. Located to the west of Seattle, this roundabout is in a mostly residential neighborhood with two small shopping centers and a high school in the vicinity. A logging industry close by results in a steady st ream of logging truck s utilizing this roundabout. The NCHRP 3 65 team provided a total of six hours of video over two days for this site. Data for this site w ere extracted from the second and third hour on the first day of filming. Using the data extract ion techniques describe d previously, these two hours of data from the three appr oa ch Port Orchard site were extracted. First, the origin destinations were completed and are shown in Table 3 1 while Table 3 2 shows the entering and conflicting flow for ea ch approach. Approach delay, average queue, and maximum queue were extracted next and are presented in Table 3 3 These approaches were not at capacity for any of the fifteen minute periods analyzed so the minimum delay and queue length are all zero and were therefore not included in the table.

PAGE 41

41 Figure 3 2 Overview of the Port Orchard, WA s ite ( Source: Google Earth). Omnidirectional Camera Northern Approach Southern Approach Eastern Approach Figure 3 3 Camera v iews of the Port Orchard WA site.

PAGE 42

42 Table 3 1 Origin Destinations for the Port Orchard, WA s ite Origin Approach: Northern Southern Eas tern Destination Approach : Eastern Southern Northern Eastern Southern Northern Period 1 100 72 65 64 64 94 Period 2 111 76 60 51 56 83 Period 3 83 61 73 79 94 99 Period 4 103 79 74 54 57 103 Period 5 98 66 74 58 61 88 Period 6 89 91 58 59 75 83 Per iod 7 92 74 62 35 63 88 Period 8 84 54 61 57 58 94 Table 3 2 Entering and c onflicting f low s ummary for the Port Orchard, WA s ite Northern Approach Eastern Approach Southern Approach Time Interval Enter ing Flow Conflicting Flow Entering Flow Conflicting Flow Entering Flow Conflicting Flow Period 1 172 64 158 65 129 100 Period 2 187 56 139 60 111 111 Period 3 144 94 193 73 152 83 Period 4 182 57 160 74 128 103 Period 5 164 61 149 74 132 98 Period 6 180 75 158 58 117 89 Period 7 166 63 151 62 97 92 Period 8 138 58 152 61 118 84

PAGE 43

43 Table 3 3 Summary of e xtracted f ield d ata from the Port Orchard, WA s ite Northern Approach Time Interval Period 1 Period 2 Period 3 Period 4 Period 5 Period 6 Period 7 Period 8 Average Delay (s) 00:09.1 00:10.3 00:12.1 00:09.4 00:07.7 00:12.6 00:22.6 00:08.6 Maximum Delay (s) 00:38.5 00:31.7 00:42.4 00:34.0 00:29.2 00:41.9 01:07.9 00:38.5 Average Queue (veh) 3.1 3.6 3.2 3.5 2.8 4.0 5.8 2.2 Maximum Queue (veh) 9.0 11.0 10.0 8.0 9.0 8.0 14.0 8.0 Eastern Approach Time Interval Period 1 Period 2 Period 3 Period 4 Period 5 Period 6 Period 7 Period 8 Average Delay (s) 00:04.2 00:03.0 00:06.7 00:06.6 00:06.4 00:03.1 00:08.5 00:05.2 Maximum Delay (s) 00:16.8 00:15.7 00:28.6 00:20.8 00:18.0 00:15.2 00:29.7 00:19.7 Average Queue (veh) 1.2 0.9 2.3 1.9 1.7 1.2 2.0 1.4 Maximum Queue (veh) 4.0 4.0 5.0 5.0 5.0 4.0 6.0 6.0 Southern Approach Time Interval Period 1 Period 2 Period 3 Period 4 Period 5 Period 6 Period 7 Period 8 Average Delay (s) 00:05.4 00:07.3 00:04.0 00:06.8 00:06.5 00:05.2 00:06.3 00:06.4 Maximum Delay (s) 00:21.6 00:38.1 00:17.5 00:23.3 00:29.9 00:18.1 00:29.5 00:24.6 Average Queue (veh) 1.5 1.7 1.7 1.8 1.8 1.4 1.3 1.4 Maximum Queue (veh) 6.0 5.0 4.0 5.0 7.0 5.0 5.0 5.0

PAGE 44

44 Four Approach Roundabout: Lothian, Maryland The second site used for data extraction was a four approach roundabout located in Lothian, Maryland. This site provide d geographic diversity to the W est Co asts Port Orchard site, and is a more traditional four approach roundabout This site has full coverage of three approaches and an omnidirectional camera to capture the circulating traffic. The fourth approach was not filmed due to the limite d number of DVD recorders and digital cameras. However, this did not result in a loss of critical information because the origin -destinations can still be obtained from the omnidirectional camera and the max imum entering flow during a fifteen minute period for this approach was only thirteen vehicles. Figure 3 4 s h ows an overview of this site and Figure 3 5 shows the camera coverage available. This site is located east of Washington D.C. in a rural area of Maryland. A total of three hours of video from t his site was recorded by the NCHRP 3 65 research team and two hours of data were extracted for use in this research. A summary of the origin destinations and the conflicting vs. entering flow is presented in Table 3 4 and Table 3 5 Average delay and que ue along with maximum delay and queue are presented in Table 3 6 Similarly to the Port Orchard site, none of the fifteen minute periods for these approaches had continuous traffic resulting in a minimum delay and queue of zero for each of the periods red uced.

PAGE 45

45 Figure 3 4 Overview of the Lothian, MD s ite ( Source: Google Earth) Omnidirectional Camera Eastern Approach Northern Approach Southern Approach Figure 3 5 Camera Views of the Lothian, MD s ite

PAGE 46

46 Table 3 4 Origin Destinations for the Lothian, M D s ite Origin Approach Northern Southern Eastern Western Destination Approach Eastern Southern Western Western Northern Eastern Southern Western Northern Northern Eastern Southern Period 1 112 50 9 1 46 3 5 4 85 6 2 2 Period 2 102 39 7 3 39 4 3 5 115 4 2 1 Period 3 118 47 5 0 45 9 7 7 80 4 2 3 Period 4 104 49 9 3 40 2 7 3 96 4 7 0 Period 5 125 58 4 7 59 7 3 5 90 6 3 4 Period 6 145 45 3 1 50 9 5 3 92 2 7 3 Period 7 118 64 9 8 48 8 1 7 93 9 2 2 Period 8 144 59 6 5 49 5 5 8 101 4 5 3 Table 3 5 E ntering and c onflicting f low s ummary for the Lothia n, M D s ite Northern Approach Eastern Approach Southern Approach Western Approach Time Interval Entering Flow Conflicting Flow Entering Flow Conflicting Flow Entering Flow Conflicting Flow Entering Flow Conflicting Flow Period 1 171 10 94 53 50 120 10 1 67 Period 2 148 11 123 46 46 108 7 144 Period 3 170 14 94 49 54 124 9 172 Period 4 162 13 106 47 45 115 11 160 Period 5 187 15 98 72 73 134 13 186 Period 6 193 9 100 53 60 154 12 195 Period 7 191 16 101 65 64 129 13 183 Period 8 209 18 114 58 59 153 12 208

PAGE 47

47 Table 3 6 Summary of e xtracted f ield d ata from the Lothian, MD s ite Northern Approach Time Interval Period 1 Period 2 Period 3 Period 4 Period 5 Period 6 Period 7 Period 8 Average Delay (s) 00:03.6 00:03.1 00:03.8 00:06.3 00:06.3 00:05.6 00:08.5 00:11.8 Maximum Delay (s) 00:17.0 00:10.6 00:12.6 00:28.4 00:27.2 00:17.3 00:32.4 00:41.4 Average Queue (veh) 1.4 1.2 1.5 2.0 2.3 2.2 2.7 3.8 Maximum Queue (veh) 5.0 4.0 6.0 9.0 9.0 7.0 9.0 10.0 Eas tern Approach Time Interval Period 1 Period 2 Period 3 Period 4 Period 5 Period 6 Period 7 Period 8 Average Delay (s) 00:10.5 00:07.4 00:05.7 00:04.5 00:04.7 00:08.9 00:08.1 00:06.6 Maximum Delay (s) 00:40.1 00:27.5 00:29.3 00:19.7 00:16.8 00:34.6 00:27.3 00:21.7 Average Queue (veh) 1.6 1.7 1.1 1.1 1.0 1.6 1.6 1.6 Maximum Queue (veh) 9.0 7.0 5.0 8.0 6.0 7.0 9.0 7.0 Southern Approach Time Interval Period 1 Period 2 Period 3 Period 4 Period 5 Period 6 Period 7 Period 8 Average Delay (s) 00:08.8 00:07.3 00:07.9 00:06.6 00:07.6 00:11.6 00:09.7 00:09.0 Maximum Delay (s) 00:44.1 00:22.1 00:29.6 00:29.1 00:45.9 00:42.9 00:44.4 00:34.7 Average Queue (veh) 0.7 0.7 0.7 0.5 1.0 0.9 1.0 0.9 Maximum Queue (veh) 5.0 4.0 4.0 4.0 4.0 5.0 4.0 4.0

PAGE 48

48 CHAPTER 4 CORSIM SIMULA TION Introduction This chapter describes the replication of traffic operations at the two roundabouts using CORSIM. It demonstrates the method used for ensuring accurate reproduction of the origindestinations through each roundabout, and an analysis of t he simulation output results. Network Implementation Network implementation of the two study roundabouts requires a number of changes to both global CORSIM parameters and NETSIM specific parameters. These changes along with how the two study roundabout ne tworks were constructed in CORSIM is the topic of this section. The first step was the creation of eight periods comprising fifteen minutes each. Next the vehicle entry headway distribution was modified to more closely resemble the arrivals at each rou ndabout approach in the field. CORSIM offers three types of headway distributions: constant headway (default), normally distributed, and an Erlang distribution with different shape pa rameters ranging from 1 to 9. To determine the best fit, field observed headway distributions were plotted and compared to CORSIMs distribution for constant headways, normally distributed headways, and each of the nine different shapes of an Erlang distribution These comparisons were conducted in the statistical software Minitab (Minitab, 2008) where testing for goodness of fit showed that a n Erlang 1 distribution (exponential) provided the best fi t Figure 4 1 and Figure 42 show the plots of the actual headway distribution and the Erlang 1 distribution for the two sites. In addition to the arrival distributions available in CORSIM, the field data were also compared to other arrival distributions including Weibull, Box Cox transformation lognormal, and loglogistic. None of these distributions resulted in a good fit to field data.

PAGE 49

49 Lothian Arrival Headway Distributions 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0 55.0 60.0 Arrival Headway (s) Probability Actual Distribution Erlang 1 Distribution Figure 4 1 Four a pproach Lothian, MD a rrival d istributions Port Orchard Arrival Headway Distributions 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0 55.0 60.0 Arrival Headway (s) Probability Actual Distribution Erlang 1 Distribution Figure 4 2 Three a pproach Port Orchard, WA a rrival d istributions

PAGE 50

50 With the ch anges to the network properties completed, the next task was to implement the roundabout networks. Google satellite images were imported into CORSIM and then scaled to match CORSIMs coordinate system Nodes were then placed at each approach to the round about and connected using one -way links. Each of these nodes was then connected to a dummy node upstream and a corresponding vehicle generation node (8000 node ) Figure 4 3 and Figure 4 4 show the networks used for the two study roundabouts. Figure 4 3 Port Orchard, WA CORSIM n etwork

PAGE 51

51 Figure 4 4 Lothian, MD CORSIM n etwork The next step in network implementation was the placement of the yield contro ls on the external approaches to the roundabout. CORSIM treats yield signs similarly to stop signs with a vehicle slowing down as it appro aches the sign. The difference is that at a yield sign, the vehicle begins looking for gaps within one second of the yield line and will continue through without stopping if an acceptable gap is found. To ensure yield sign placement within CORSIM replicated roundabout operations the best, the roundabouts were also tested with no control and with stop signs. Stop signs predictably drove up control delay to unrealistic numbers while no control reduced control delay to minimal values. When stop signs or no control was used, the animation looked unrealistic for roundabout operations leading to the conclusion that yield s igns are indeed the best for replicating roundabouts.

PAGE 52

52 In addition to properly implementing the network nodes and intersection control type link lengths for the links representing the circulating roadway of the roundabout were modified CORSIM automatical ly calculates a straight line distance between nodes as a default link length which needs to be corrected to match the curved length of the actual roadway. This allows these links not only to be the correct length, but for C ORSIM to give them an accurate counterclockwise curve in the animation. Figure 4 5 and Figure 4 6 s how the CORSIM networks as displayed by an imation. Figure 4 5 Port Orchard, WA CORSIM a nimation

PAGE 53

53 Figure 4 6 Lothian, MD CORSIM a nimation The final task completed for network implementation was to adjust the gap acceptance models to match the findings of NCHRP 3 65. NCHRP 3 65 computed a critical gap value of 4.5 seconds when consi dering all observations of lags (vehicle entry with no conflicting flow) and gaps (vehicle entry with at least one conflicting vehicle). This number was used for driver type 5, which is considered an average driver. The other driver type gap acceptances were then determined using the same differences found between driver types as those of the default CORSIM values. Table 4 1 shows the default CORSIM values of gap acceptance and the modifications discussed based o n the critical gap found in NCHRP 3 65.

PAGE 54

54 Table 4 1 Acceptable g ap for right t urning v ehicles Driver Type 1 2 3 4 5 6 7 8 9 10 Default Gap Acceptance 10.0 8.8 8.0 7.2 6.4 6.0 5.6 5.2 4.8 3.6 Roundabout Gap A cceptance 8.1 6.9 6.1 5.3 4.5 4.1 3.7 3.3 2.9 1.7 Inputting Roundabout Origin-Destinations The proper implementation of origin destinations for CORSIM roundabouts is possible using conditional turn movements. This section explain s how volumes and turn m ovements were input into the simulation of the two study roundabouts. The number of vehicles on each link was th en compared between field data and CORSIM to evaluate the accuracy of the method used. The first step in accurately replicating the traffic on the two roundabout networks is in the generation of traffic at the 8000 nodes. Field data provided the entering flow on each roundabout approach for each of the eight periods collected. These values, in vehicles per fifteen minutes, were then multiplied by four to represent vehicles per hour and input into their respective periods within CORSIM. In addition to the flow for each period, the percentage of heavy vehicles was also input. Distribution of this generated traffic is the next step and is handled through conditional turn movements at each roundabout node. What makes traffic distribution at roundabouts difficult in CORSIM is its reliance on turning percentages at each node. Each node looks one node upstream and distributes the traffic arriving fro m there by user input percentages corresponding to percent going left, through, or right. Normally this works well because each node represents an isolated intersection. However, roundabouts must be modeled as a series of nodes which operate together to move traffic through the roundabout network. A vehicle can enter a roundabout approach node and circulate past one or more downstream nodes To correct this potential for error conditional turn movements must be utilized.

PAGE 55

55 An example of the method used f or origin -destination replication is provided next, for the Port Orchard site. T he coding of both relative turn volumes and conditional turn movements for the roundabout approaches assuming no u turns at the roundabout is conducted as follows : 1 External approaches were coded by inputting vehicle counts going right (exiting at the next downstream roundabout node) or through (exiting two roundabout nodes downstream). The downstream node ID for both right and through movements is the next downstream roundabout node. 2 Internal approaches were coded with left and through movements departing to the next downstream roundabout node while right movements were coded to the downstream dummy node for that approach. The relative turn volumes for left, through, and right are assigned a dummy volume of 1 since these will be overwritten by conditional turn movements anyway 3 Conditional turn movements for the internal approaches were coded with 100% of upstream left turners making a right movement, 100% of the upstream thr ough movement making a left movement, and 100% of right turners making a right movement. Origin -destinations within a four approach roundabout are more difficult to replicate in CORSIM due to the addition of the extra approach. This results in the need t o look not only two roundabout nodes upstream as in the Port Orchard site but three. The steps used for accomplishing this in the case of the Lothian four approach roundabout are : 1 External approaches were coded by inputting vehicle counts g oing left, th rough, or right at the roundabout intersection. The downstream node ID for these left, through, and right movements is still the next downstream roundabout node as at the Port Orchard site 2 Internal approaches were coded with left, through, and right dia gonal movements departing to the next downstream roundabout node while right movements were coded to the upstream dummy node for that approach. The relative turn volumes for left, through, right, and diagonal are assigned a dummy volume of 1 since these w ill be overwritten by conditional turn movements. 3 Conditional turn movements for the internal approaches were coded with 100% of upstream left turners making a through movement, 100% of the upstream through movement making a diagonal movement, 100% of right turners making a right turn movement, and 100% of upstream diagonal turns making a right turn movement. Figure 4 7 shows CORSIM s logic in tracking vehicles through the roundabout for proper origin destination replication.

PAGE 56

56 Figure 4 7 CORSIM s l ogic for c onditional t urn m ovements at r oundabouts Table 4 2 and Table 4 3 s how a comparison between the link volumes compute d by CORSIM and the actual link volumes computed from the collected data for the two study roundabouts These tables indicate that the volumes on each link are almost always within 5 vehicles of the actual counts collected in the field data. While some l inks experience percent differences in excess of 10% during a single period, the other seven periods are very accurately replicated. From these tables it can be concluded that link volumes are replicating the field data well ensuring realistic origin -dest inations at the two study roundabouts.

PAGE 57

57 Table 4 2 Link v olume c omparison at the Port Orchard s ite Time Period Link 1 -2 2 -3 3 -1 1 -7 2 -5 3 -4 5 -2 4 -3 7 -1 1 CORSIM 223 219 237 134 162 153 158 171 120 Field Data 229 223 236 136 164 159 158 172 129 Percent Diff. 2.6% 2.0% 0.4% 1.8% 1.2% 3.6% 0.3% 0.5% 7.0% 2 CORSIM 188 181 245 132 146 123 138 188 75 Field Data 222 199 243 132 162 143 139 187 111 Percent Diff. 16.6% 9.6% 1.0% 0.3% 10.7% 15.1% 0.7% 0.6% 38.6% 3 CORSIM 232 265 241 161 156 168 190 145 152 Field Data 235 266 238 155 162 172 193 144 152 Percent Diff. 1.5% 0.2% 1.4% 3.8% 4.1% 2.2% 1.6% 0.4% 0.3% 4 CORSIM 236 235 241 136 163 175 163 181 130 Field Data 231 234 239 136 157 177 160 182 128 Percent Diff. 2.1% 0.5% 0.9% 0.2% 3.9% 1.0% 1.8% 0.7% 1.3% 5 CORSIM 233 228 226 123 156 169 149 167 131 Field Data 230 223 225 127 156 162 149 164 132 Percent Diff. 1.5% 2.2% 0.3% 3.2% 0.3% 4.5% 0.2% 1.9% 0.5% 6 CORSIM 206 216 253 164 147 142 158 178 117 Field Data 206 216 255 166 148 141 158 180 117 Percent Diff. 0.1% 0.1% 0.9% 1.2% 0.6% 0.8% 0.3% 1.1% 0.1% 7 CORSIM 192 215 231 137 130 149 152 166 98 Field Data 189 213 229 137 127 150 151 166 97 Percent Diff. 1.8% 1.0% 1.0% 0. 1% 1.9% 0.9% 0.9% 0.2% 0.9% 8 CORSIM 202 214 198 112 140 157 151 139 117 Field Data 202 213 196 112 141 155 152 138 118 Percent Diff. 0.0% 0.7% 0.8% 0.4% 0.6% 1.0% 0.5% 0.7% 1.0%

PAGE 58

58 Table 4 3 Link v olume c omparison at the Lothian s ite Time Period Link 1 -2 2 -3 3 -4 4 -1 1 -7 2 -6 3 -5 4 -8 7 -1 6 -2 5 -3 8 -4 1 CORSIM 168 145 178 175 56 117 137 13 49 94 170 10 Field Data 170 147 181 177 57 117 137 14 50 94 171 10 Percent Diff. 0.9% 1.2% 1.7% 0.9% 2.5% 0.2% 0.0% 9.0% 2.6% 0.5% 0.7% 3.0% 2 CORSIM 156 167 163 154 44 111 155 15 46 122 151 6 Field Data 152 167 157 151 43 108 158 13 46 123 148 7 Percent Diff. 2.9% 0.3% 3.8% 1.8% 1.4% 2.8% 2.2% 15.6% 0.2% 1.0% 1.9% 15.4% 3 CORSIM 174 143 183 18 3 64 124 128 9 56 94 168 10 Field Data 178 143 184 181 57 129 129 12 54 94 170 9 Percent Diff. 2.4% 0.2% 0.8% 1.0% 11.1% 3.6% 0.6% 25.4% 2.9% 0.2% 1.5% 10.5% 4 CORSIM 166 155 177 174 54 116 139 15 45 106 162 11 Field Data 160 153 175 171 56 113 140 15 45 106 162 11 Percent Diff. 3.5% 1.3% 1.4% 1.5% 3.6% 2.4% 0.6% 0.0% 0.4% 0.0% 0.2% 0.9% 5 CORSIM 205 169 206 200 65 136 150 18 70 99 187 12 Field Data 207 170 202 199 65 135 155 16 73 98 187 13 Percent Diff. 1.0% 0.9% 2.0% 0.6% 0.2% 0.4% 3.2% 10. 7% 3.8% 1.3% 0.2% 10.5% 6 CORSIM 217 157 204 210 55 161 144 7 62 100 192 13 Field Data 214 153 202 207 53 161 144 7 60 100 193 12 Percent Diff. 1.3% 2.4% 0.9% 1.3% 4.2% 0.1% 0.1% 2.9% 3.1% 0.0% 0.4% 5.7% 7 CORSIM 194 164 207 198 65 131 149 22 62 101 191 13 Field Data 193 166 207 196 67 128 150 24 64 101 191 13 Percent Diff. 0.7% 1.0% 0.2% 0.8% 2.9% 2.3% 0.8% 9.2% 2.9% 0.2% 0.1% 3.9% 8 CORSIM 214 170 229 225 70 156 150 17 59 113 209 13 Field Data 212 172 227 220 67 154 154 19 59 114 209 12 Per cent Diff. 0.8% 1.2% 0.7% 2.0% 4.2% 1.2% 2.7% 10.5% 0.3% 1.2% 0.2% 8.8%

PAGE 59

59 CORSIM Simulation Output Analysis With the completion of network implementation and origin destination assignments for the Port Orchard and Lothian roundabout s the roundabout s w ere simulated ten times using a different random seed number fo r each simulation. These ten simulations were tested using Equation 4 1 where ten runs were determined to be sufficient given a tolerance of 1.25 seconds and a confidence interval of 90%. 2 2 2s e z N (4 1) where, N = number of runs z = confidence interval e = tolerance s = standard deviation The average across these ten simulations for control delay, average queue, and maximum queue was then taken and compared to the field data for each period. These comparisons were then used to calibrate the models to the collected field data. Work on this research concluded that mean startup delay is the only link specific calibration parameter that can be used for roundabouts in the softwares current version Mean startup delay is the drivers response time to the ability to complete their desired maneuver based on the control device. At a roundabout it can be used in calibration by increasing or decreasing the value from the default of 2.0 seconds. An increase in mean startup delay will increase control delay by slowing down the discharge into the roundabout. Decreasing mean startup delay allows a faster discharge into the roundabout thereby decreasing control delay. This parameter was adjusted for the two simulated roundabouts until the simulated control delays

PAGE 60

60 across all the time periods for each approach matched as closely as possible to the control delays observed in the field. Table 4 4 and Table 4 5 represent the best results obtained by this calibration for the Port Orchard and Lothian site respectively. The Port Orchard site, Table 4 4 shows control delay differe nce s between CORSIMs output and field data varies from 0 to 1 3 .9 seconds across the eight observed periods The biggest difference of 13.9 seconds occurs on the northern approach in the seventh period when the actual control delay is ten seconds per vehi cle higher than any other period observed during data collection. The next highest control delay difference is only 3.6 seconds. While 3.6 seconds may seem like a small number, it is best to look at the percent difference in determining CORSIMs ability to accurately simulate the roundabout. Percent differences on control delay range from 0.1% to 88.4% (13.9 second difference) or 61.7% (3.6 second difference). The wide range of differences indicates that CORSIM is having trouble accurately predicting co ntrol delay across all the observed periods. Average queue is under predicted for this site with percent differe nces between 60.3% and 114.2%. These large differences indicate a systematic error for average queue since none of the values predicted by CORS IM were close t o those measured in the field. Maximum queue is generally over predicted by CORSIM for this site with percent dif ferences between 1.4% and 71%. This indicated that some periods are being well predicted while others are not. In comparison t o the Port Orchard site, the Lothian four approach roundabout was more closely replicated in CORSIM as shown by Table 4 5 Control delay differences between field data and simulation range from 0.1 to 4.6 seconds or 1.1 to 48.7 percent difference across the eight simulated periods. These numbers fluctuate between over prediction and under prediction indicating a period specific calibration parameter is needed for better control.

PAGE 61

61 Average queue as with the Lothian, MD site is under predicted for this site with percent differences ranging from 58.1% to 168.9%. These values support the Lothian values in indicating there is a systematic problem occurring between CORSIMs calculation of control delay and measured field values. Maximum queue for this site is mostly under predicted with percent differences ranging from 1.1% to 48.3%. As with the Lothian four -approach site, some periods were accurately replicated while others were not. Even though some of these percent d ifferences for maximum queue were high, the biggest difference between observed maximum queue and CORSIM maximum queue was only two vehicles. Table 4 4 Comparison between CORSIM and f ield d ata for the Port O rchard, WA s ite Control Delay (s/veh) Time Period Eastern Approach Northern Approach Southern Approach CORSIM Actual Percent Diff. CORSIM Actual Percent Diff. CORSIM Actual Percent Diff. 1 5.1 4.2 18.7% 8.7 9.1 4.3% 6.8 5.4 23.7% 2 3.2 3.0 7.4% 10.3 10.3 0.1% 5.8 7.3 23.6% 3 8.2 6.7 20.0% 10.8 12.1 11.8% 7.6 4.0 61.7% 4 6.6 6.6 0.4% 9.8 9.4 4.6% 7.7 6.8 12.4% 5 5.6 6.4 13.5% 7.8 7.7 0.9% 7.2 6.5 10.5% 6 5.0 3.1 47.5% 12.5 12.6 0.7% 6.0 5.2 14.3% 7 5.7 8.5 40.2% 8.7 22.6 88.4% 5.7 6.3 9.9% 8 5.0 5.2 3.2% 6.1 8.6 33.4% 5.3 6.4 18.4% Cumulative Average Queues (veh) Time Period Eastern Approach Northern Approach Southern Approach CORSIM Actual Percent Diff. CORSIM Actual Percent Diff. CORSIM Actual Percent Diff. 1 0.6 1.2 60.3% 1.1 3.1 91.4% 0.5 1.5 92.0% 2 0.5 1.1 77.5% 1.3 3.3 85.1% 0.4 1.6 114.2% 3 0.8 1.5 61.4% 1.4 3.3 82.0% 0.5 1.6 100.4% 4 0.8 1.6 62.5% 1.4 3.3 82.2% 0.6 1.6 97.9% 5 0.8 1.6 66.2% 1.3 3.2 84.9% 0.6 1.7 97.2% 6 0.8 1.5 65.6% 1.4 3.3 80.4% 0.6 1.6 98.2% 7 0.8 1.6 69.8% 1 .4 3.7 90.7% 0.5 1.6 100.3% 8 0.8 1.6 70.8% 1.3 3.5 92.3% 0.5 1.5 102.2% Cumulative Maximum Queues (veh) Time Period Eastern Approach Northern Approach Southern Approach CORSIM Actual Percent Diff. CORSIM Actual Percent Diff. CORSIM Actual Percent Dif f. 1 6.9 4.0 53.2% 8.1 9.0 10.5% 5.6 6.0 6.9% 2 7.5 4.0 60.9% 11.4 11.0 3.6% 6.2 6.0 3.3% 3 9.6 5.0 63.0% 11.8 11.0 7.0% 8.4 6.0 33.3% 4 10.3 5.0 69.3% 12.2 11.0 10.3% 8.9 6.0 38.9% 5 10.5 5.0 71.0% 12.7 11.0 14.3% 9.3 7.0 28.2% 6 10.5 5.0 71.0% 13.7 11.0 21.9% 9.5 7.0 30.3% 7 10.7 6.0 56.3% 13.7 14.0 2.2% 9.7 7.0 32.3% 8 11.2 6.0 60.5% 13.8 14.0 1.4% 9.8 7.0 33.3%

PAGE 62

62 Table 4 5 Comparison between CORSIM and f ield d ata for the Lothian, MD s ite Control Delay (s/veh) Time Period Southern Approach Eastern Approach Northern Approach CORSIM Actual Percent Diff. CORSIM Actual Percent Diff. CORSIM Actual Percent Diff. 1 7.0 8.8 22.7% 6.5 10.5 46.6% 4.3 3.6 17.6% 2 6.8 7.3 6.7% 7.3 7.4 0.8% 4.4 3.1 33.7% 3 7.7 7.9 2.3% 5.7 5.7 0.4% 5.1 3.8 28.3% 4 6.8 6.6 2.5% 6.0 4.5 28.5% 5.1 6.3 21.7% 5 10.7 7.6 34.2% 7.7 4.7 47.8% 6.2 6.3 1.9% 6 11.7 11.6 1.1% 6.2 8.9 35.5% 5.1 5.6 8.5% 7 8.9 9.7 8.9% 6.6 8.1 20.8% 6.4 8.5 28.7% 8 9.8 9.0 8.8% 7.3 6.6 10.6% 7.2 11.8 48.7% Cumulative Average Queues (veh) Time Period Southern Approach Eastern Approach Northern Approach CORSIM Actual Percent Diff. CORSIM Actual Percent Diff. CORSIM Actual Percent Diff. 1 0.3 0.7 90.4% 0.4 1.6 112.0% 0.1 1.4 168.9% 2 0.3 0.7 91.0 % 0.6 1.6 97.7% 0.1 1.3 167.3% 3 0.3 0.7 84.1% 0.5 1.4 97.4% 0.2 1.4 160.1% 4 0.3 0.7 80.5% 0.5 1.4 94.2% 0.2 1.5 159.2% 5 0.4 0.7 65.1% 0.5 1.3 84.8% 0.2 1.7 153.1% 6 0.4 0.8 56.4% 0.5 1.3 89.5% 0.2 1.8 155.2% 7 0.4 0.8 58.3% 0.5 1.4 91.6% 0.3 1.9 152.3% 8 0.4 0.8 58.1% 0.5 1.4 90.4% 0.3 2.1 150.6% Cumulative Maximum Queues (veh) Time Period Southern Approach Eastern Approach Northern Approach CORSIM Actual Percent Diff. CORSIM Actual Percent Diff. CORSIM Actual Percent Diff. 1 3.6 5.0 32.6% 5.5 9.0 48.3% 4.4 5.0 12.8% 2 4.0 5.0 22.2% 7.2 9.0 22.2% 5.2 5.0 3.9% 3 4.7 5.0 6.2% 7.9 9.0 13.0% 6.3 6.0 4.9% 4 5.1 5.0 2.0% 8.0 9.0 11.8% 7.2 9.0 22.2% 5 6.4 5.0 24.6% 8.3 9.0 8.1% 7.9 9.0 13.0% 6 6.8 5.0 30.5% 8.6 9.0 4.5% 7.9 9.0 13.0% 7 6.8 5.0 3 0.5% 8.7 9.0 3.4% 9.1 9.0 1.1% 8 6.9 5.0 31.9% 8.7 9.0 3.4% 11.0 10.0 9.5% CORSIM Implementation Summary and Conclusions Simulation of roundabouts in CORSIM begins with proper implementation of the network. There were four keys to the successful implem entation of the study roundabout s in to CORSIM. First, vehicle entry headways were changed to an Erlang 1 distribution because it was the closest match to the field observed headways of the available distributions Secondly i ndividual nodes were placed a t each approach to the roundabout and connected using a series of one -way links. Next gap acceptance parameters for right turners were lowered to coincide with

PAGE 63

63 the critical gap headway determined by NCHRP 3 65. Finally, mean startup delay was used as a calibration parameter to ensure the smallest differences between CORSIMs estimation of control delay and those observed in the field across all of the available periods The next step for implementation of the two study roundabouts was the use of conditio nal turn movements to replicate accurately origin -destinations through the roundabout. The interconnected nature of roundabout nodes makes it necessary to look more than one node in advance for turning movements. The method for coding the turn movement a nd conditional turn movement percentages for a three approach and four approach roundabout has been discussed previously and a s demonstrated by Table 4 2 and Table 4 3 t hes e origin -destination methods are effective. With the roundabout networks built the final task was to simulate the roundabouts and compare the field data to CORSIMs output. Control delay comparisons show a wide range of errors for CORSIMs prediction of control delay at both roundabout sites. Table 4 4 and Table 4 5 shows that CORSIM is both over predicting and under predicting the control delay at these sites. This over /under predicting indicates that the current version of CORSIM can model roundabouts in a general way but needs to have more control over individual periods to account for specific driver behavior during different periods A good example of situations where CORSIMs roundabout modeling needs improvement is at the northern approach to the Port Orchard roundabout during period seven. The turning volumes during this period are roughly equivalent to all the other periods yet the control delay is ten seconds per vehicle higher than any other period collected. Observation of the video for this period shows a number of timid roundabout users at the beginning of traffic platoons. These drivers stop of yield unnecessarily on the approach and since they are at th e

PAGE 64

64 beginning of a platoon they increase the control delay for all the following vehicles. While CORSIM can account for different driver types, the specific arrival time of these different drivers cannot be controlled. What is needed is the ability to contr ol departures on a period specific basis. This would allow CORSIM to take into account how driver behavior varies across periods. For example, a period with more timid drivers would require a higher gap acceptance value. Similarly, a period with more ag gressive drivers would need a lower gap acceptance value. This ability would allow for more accurate and realistic simulation of a roundabout throughout all periods. Average queue was the least well predicted of the three performance measures. At both si tes this value was underestimated by between 58% and 160% indicating more of a systematic error. The answer for this systematic error lies in a discrepancy in CORSIMs definition of a queued vehicle and that of the field collected data. CORSIM defines a queued vehicle as: having an acceleration of less than 2 ft/s/s and a velocity less than 9 ft/sec If the velocity is between 3 and 9ft/s it is considered queued every other second and if the velocity is less than 3 ft/sec it is considered queued every se cond. Field data was calculated by counting the number of cars every 15 seconds between the upstream Z line and entry into the roundabout who experience control delay. The field method would count vehicles slowing down while CORSIM would not. This wou ld result in the under estimation seen by the CORSIM results since field queue s would typically be greater. Maximum queue was the final performance measure compared between CORSIM and field values. Like control delay, maximum queue was both over and under predicted at both sites. The differences range from between 1.1% to 69.3% across both sites. These errors can be attributed to two things. First, there is error involved once again due to the definition of a

PAGE 65

65 queued vehicle as discussed previously. The numbers for maximum queue were better simulated between CORSIM and field data This results because during a maximum queue situation no other vehicles are slowing down for the queue so the field data would not be counting any excess vehicles and be clos er to the CORSIM predicted values. The last area with potential for errors comes from maximum queue being a cumulative property across all the periods. This means the comparison is highly dependent on which period the maximum queue arrives. For example, CORSIM may predict a maximum queue in the first period while it really doesnt occur until the sixth period. This will lead to periods one through five having significant differences. Added to this is the importance of arrivals to maximum queue and as indicated in Figure 4 1 and Figure 4 2 CORSIM s arrival distribution does not perfectly match that observed in the field data. Even though control delay varied widely between CORSIMs output and the collected field data for individual periods, it was well predicted on average. Table 4 6 s hows the average values for CORSIMs output and the field data across the eight time periods for the three approach roundabout. Two of the approaches have percent differences within 10%, while the third approach has a lar ger percent difference due to the large actual control delay in the sev enth period. Table 4 7 demonstrates the same for the four approach roundabout in Lothian, MD. These tables demonstrate that CORSIM, on average, can predict control delay at roundabout s reasonably well. It is only on a period specific basis that it loses accuracy. Table 4 6 Average control delay values across the 8 time periods for the Port Orchard, WA s ite Approach Avg. CORSIM Value Av g. Field Value Percent Diff. Eastern 5.55 5.46 1.59% Northern 9.34 11.55 21.18% Southern 6.51 5.99 8.40%

PAGE 66

66 Table 4 7 Average control delay values across the 8 time periods for the Lothian, MD s ite Approac h Avg. CORSIM Value Avg. Field Value Percent Diff. Southern 8.68 8.56 1.31% Eastern 6.66 7.05 5.65% Northern 5.48 6.13 11.21%

PAGE 67

67 CHAPTER 5 RESEARCH CONCLUSIONS Introduction The results of this research show that CORSIM in its current version does not replicate roundabout operations across all the observed periods However, many of the periods had differences of less than 20% for control delay. A series of changes are recommended for CORSIM as a result of this research These recommended changes and recommendat ions for future roundabout work in CORSIM are presented in this chapter. Recommended CORSIM Changes The literature review and resulting research for this thesis have identified a number of improvements that can be implemented in CORSIM. These can be split into two distinct categories The first category is the overall user friendliness and animation of roundabout implementation. The second area involves more specific adjustments to the way traffic is simulated at a roundabout. T he se recommended changes are summarized in this section. User Friendliness and Animation Changes Roundabout coding in CORSIM requires only a small number of steps, however, these steps are prone to input errors. The critical nature of each of these steps and the potential for inp ut errors means a roundabout coder needs to be careful during the coding process. Through the testing of the two study roundabout s in CORSIM, it was determined that circulating speed has a significant impact on the simulations output for operational chara cteristics of the roundabout. NCHRP report 572 reports that current speed prediction methods for predicting 85thpercentile circulating speeds appear to be reliable (Rodegerdts et al 2007). Since these methods have been proven reliable, the first rec ommendation is to have CORSIM automatically calculate a circulating speed as a default based on Equation 5 1

PAGE 68

68 ) ( 15 f e R V (5 1) w here V = speed (mph) R = vehicle path radius (ft) e = superelevation (ft/ft) f = side friction factor This allows a roundabout modeler to have confidence in the estimated speed when trying different roundabout designs while still allowing a modeler to input their own values based on available field data. Origin -destination co ding is another area that can be improved for roundabouts. While the final method devised for origin destination implementation has only three steps, it is critical for these steps to be performed correctly. The ability to select and group nodes together to form a roundabout would be useful for the coding of origin -destinations Once grouped, CORSIM could automatically fill in the conditional turn movement information and ask the user to define the turning volumes at each approach to the roundabout inter section. This is already being accomplished during the conversion from HCS+ to CORSIM so it should be a relatively minor change for it to happen within CORSIM itself. This will ensure accurate reproduction of traffic on each link by reducing the chance f or user coding errors The simulation of internal links representing the circulating roadway of a roundabout is another area for improvement These links are currently coded on an individual basis to simulate the curved length between two roundabout nodes and then given a counter clockwise curvature for a circular shape to appear in the animation file. If the coding of these internal link lengths could be accomplished automatically by the program based on inscribed diameter and approach node placement it w ould reduce coding errors while making roundabout implementation easier.

PAGE 69

69 This could be done in a similar method to origin -destinations and circulating speed whereby nodes can be grouped together and speed, conditional turn movements, and link lengths are all automatically coded based on a single menu of user inputs. The final area of recommended improvements to user friendliness and animation is how the roundabout looks in the animation. Roundabouts currently look fairly realistic when view ed in the ani mation except for one area. The approach links to the roundabout s d o not have the characteristic splitter island and flare associated with todays modern roundabout. Instead, these intersections look like 90 degree T intersections. If the animation ca n be modified to incorporate these splitter islands into roundabouts, then they would look more realistic in the animation Traffic Simulation Changes The second area of CORSIM improvement recommendations are modifications to how vehicular traffic is simul ated at a roundabout and the network as a whole. These c hanges are like ly to be much more difficult to include in a future release because they involve issues of approach and time specific modifications to the model These changes include the incorporati on of follow up time and critical gap into each roundabout approach along with the ability for the user to provide/change the definition of a queued vehicle CORSIMs definition of a queued vehicle having less than 2 ft/s/s acceleration and a velocity less than 9 ft/s works well at most intersections types. However, roundabout queues act like an accordion with large variances in the acceleration and speed of queued vehicles based on the number of vehicles entering the roundabout. When a roundabout approac hs conflicting flow rate drops to zero all vehicles queued will begin to discharge. This will then increase both their speed and acceleration as they try to close gaps between the vehicles in front of them This will result in vehicles having substantia l acceleration and speed characteristics while still being part

PAGE 70

70 of the approach queue. CORSIM does not consider this phenomenon when predicting the queue. What is needed is the ability for the user to provide a definition of a queued vehicle so that this e ffect may be captured. This will also allow a modeler to easily compare field data to CORSIM outputs for future studies. CORSIMs biggest issue when it comes to roundabout modeling is its inability to accurately simulate control delay across all periods CORSIM provides mean startup delay as the only calibration parameter. While this parameter should have an e ffect, it should be a minor one when compared to follow up time and critical gap. Roundabouts have wide variability in control delay even across similar traffic flows as demonstrated by the data collection for this research. This indicates a wide variability in critical gap and follow up time in the real world which cannot be simulated by CORSIM because gap acceptance is a global parameter and follow up time is not incorporated into the program. Changing roundabout gap acceptance from a global parameter to an approach/time specific parameter for roundabouts has two major benefits. First, gap acceptance at a roundabout is different from other int ersection types and require s a reduction in the default values for CORSIM which are based on right turn on red at signalized intersections If this research had modeled roundabouts as part of a network, the necessary changes to gap acceptance for roundabo uts would have been applied to other intersections types and would have affected the outputs. The second benefit is the increased ability to model gap acceptance as a function of driver variability throughout different periods. Specific incorporation of follow up time into the gap acceptance parameters at roundabout approaches is the final adjustment needed to the traffic model for accurate simulation of roundabouts. Si nce follow up time is one of the two main factors affecting roundabout

PAGE 71

71 operations, addition of the ability t o code approach and time specific follow -up time will allow for more robust roundabout modeling capabilities. It is especially important as a calibration parameter as it greatly affects an approaches entering flow when no conflictin g vehicles are present. Summary of Recommendations CORSIM has all of the basics necessary for proper roundabout simulation. The incorporation of the following modifications will utilize these basics allowing CORSIM to become first rate in roundabout micro -simulation. 1 Roundabout nodes should be able to be grouped. 2 Circulating speed should be automatically coded based on radius and super elevation. 3 Conditional turn movements should automatically be coded once nodes are grouped as a roundabout. 4 Internal link s are automatically coded for a counterclockwise curvature and a radius to match that of the roundabout being simulated. 5 Incorporation of splitter islands into the animation 6 Ability for the user to define the definition of a queued vehicle. 7 Approach and time specific gap acceptance at roundabout approaches. 8 Approach and time specific follow up time at roundabout approaches. The addition of these recommended improvements for user friendliness, animation, and incorporation of follow up time and critical ga p to the approaches will result in a reliable easy to use roundabout micro-simulator. A recommendation for the final form CORSIM could use for future roundabout coding is: M ultiple nodes are grouped as a roundabout. Once grouped, the software seeks inp uts for inscribed diameter, superelevation, and turning movements at each approach. Approach and time specific f ollow up time and

PAGE 72

72 critical gap will have default values based on NCHRP 3 65 with the ability to overwrite based on available field data. This information will then automatically be used to calculate the circulating speed, adjust the link length s and curvature for realistic animation, code turn movements and conditional turn movements at each approach node and adjust gap acceptance and follow up time to be approach specific If these changes are incorporated into a new release of the software, CORSIM will be one of the easiest micro -simulators available for implementing roundabouts. Additionally, t he use of approach specific follow up time and critical gap can only improve CORSIMs current ability to model roundabout performance measures. Recommendations for Future Research There are a number of areas that need future research involving CORSIM roundabout implementation. These include the implem entation of a roundabout as part of a network investigation of CORSIMs ability to model two lane roundabouts, and a study of the affects of arrival distribution on roundabout operational parameters. These t hree areas could not be addressed by this resea rch either due to lack of data or time constraints, or software limitations in CORSIMs current release. This study only investigated a roundabout modeled as a single isolated intersection rather than as part of a whole transportation network. The result s of this study showed that roundabout delay could be estimated to within a 50% difference for control delay with more than half of the periods having percent differences less than 20%. Some of these large differences are not ideal for a microscopic analy sis of a single roundabout. However, future research may show that when implemented as part of a large network, a roundabouts impact to the network on a macroscopic level can be accurately simulated. Due to time constraints, this research focused on CORSI Ms ability to model the most common modern roundabout type in the United States, the single lane roundabout. However,

PAGE 73

73 there are a growing number of multilane roundabouts and CORSIMs ability to successfully model multilane gap acceptance at roundabouts i s unknown. Once the software is updated to code approach specific gap acceptance, a similar study to this research should be completed testing CORSIMs ability to model multilane roundabouts. The final area of future roundabout research for CORSIM deals with the impact of arrival distributions. Demonstrated by Figure 4 1 and Figure 4 2 the Erlang 1 distribution which was the best fitting CORSIM distribution, does not fit the field data well. A number of other arrival distributions were tested with none fitting better than the Erlang 1. Future research into the impacts of arrival distribution on operational parameters would be useful to find out how critical these distributions are to the outcome However, the recommended changes to the traffic model s hould be fixed first. Once these have been adjusted to allow for realistic roundabout entries, the full impact of arrival distribution on control delay and queue can be determined.

PAGE 74

74 APPENDIX INTERIM GUIDE TO ROU NDABOUT MODELING IN CORSIM The recommendatio ns of the previous section will increase the reliability of roundabout modeling in CORSIM Until these changes can be implemented in a future release of the software, this section provides an interim guide to the best method for implementing a four approa ch roundabout using the current software version. 1 Begin a new TRAFED file and enter the network properties menu. 2 Adjust the time period duration to match the simulation time desired ( Figure A 1 ) Figure A 1 Step 2 A dj ust s imulation d uration 3 Under the vehicle entry headway tab change from a constant headway to an Erlang distribution with a parameter of 1 unless field data indicates otherwise This

PAGE 75

75 randomizes the vehicle arrival headways to better match the random arrivals of an isolated roundabout (Figure A 2 ) Figure A 2 Step 3 C hange a rrival d istribution 4 Place a node at each approach to the roundabout in the proper coordinates. It is recommended to scale a background image of the study site to CORSIMs coordinate system and then place a node at each roundabout approach. (Figure A 3 ) 5 Each roundabout node must then be connected via a one -way single lane link in the counterclockwise direction. (Figure A 3 ) 6 These roundabout no des must then be linked to dummy nodes and traffic generation nodes (8000 nodes) ( Figure A 3 ) 7 Place a yield sign at each external approach to a roundabout node ( Figure A 3 )

PAGE 76

76 Figure A 3 Steps 4 7 Construction of n etwork n odes 8 Each internal roundabout link must be modified in three ways. First, link length must be modified from the straight line distance between roundabout nodes to the length of the curved path between these two nodes. Seco nd, under the graphics tab select counter clockwise as the direction of curvature. This will give the link curvature in the animation making it look like a roundabout. Finally, the speed on the link must be changed to match field collected data or use th e equation from FHWAs Roundabouts: An Informational Guide (Robinson et al 2000) : (Figure A 4 ) Figure A 4 Step 8 Internal l ink m odif i cations

PAGE 77

77 9 Change the external roundabout link speed data for each approach to match t he free flow speed of the site being analyzed. 10. At the 8000 nodes generate the traffic that will be utilizing each approach link for the time periods being simulated along with inputting the percent of heavy vehicle use. (Figure A 5 ) Figure A 5 Step 10 Generation of t raffic at the 8000 n odes 11. At each roundabout node, code the external approach turn movement volumes as actual counts or percentages of vehicles going left, through or right. Assign all three of these turni ng movements to the next downstream roundabout node (Figure A 6 ) Figure A 6 Step 11 External a pproach t urn m ovement c oding

PAGE 78

78 12. The internal approach links to each roundabout node should be assigned a dummy volume of 1 for left, through, right, and right diagonal. Left, through, and right diagonal should be assigned the next downstream roundabout node for their departures while the right turn should be assigned the upstream dummy node on that approach for its departure. (F igure A 7 ) Figure A 7 Step 12 Internal a pproach t urn m ovement c oding 13. Conditional turn movements on the internal approaches must be coded as follows: 100% of upstream left turns make a through movement, 100% of the ups tream through movement makes a diagonal turn, 100% of upstream right turns make a right turn, and 100% of upstream diagonal turns make a right turn. ( Figure A 8 ) Figure A 8 Step 13 Internal a pproach c onditional t urn m o vement c oding

PAGE 79

79 14. Open the NETSIM setup under the network menu. Select the tab marked left/right turns. Right turn gap acceptance must be changed to: 8.1, 6.9, 6.1, 5.3, 4.5, 4.1, 3.7, 3.3, 2.9, and 1.7 for driver types 1 10 respectively. This a djusts the g ap acceptance values down to coincide with the roundabout critical gap value of 4.5 seconds determined by NCHRP 3 65. (Figure A 9 ) Figure A 9 Step 14 Modifications to g ap a cceptance 15. Simulate the model using multiple r uns 16. If field data exists for control delay at the roundabout being studied, use mean startup delay to calibrate the model for the closest match between CORSIM and field data. 17. E xtract the desired data from the output for reporting

PAGE 80

80 LIST OF REFERENCES Ak elik, R., Chung E. and M. Besley. 1999. Roundabouts: c apacity and p erformance a nalysis. Research Report ARR No. 321, 2nd ed. ARRB Transport Research Ltd, Australia. Akelik, R. 2003. A r oundabout c ase s tudy c omparing c apacity e stimates from a lternative a nalytical m odels. In: Proceedings of the 2 nd Urban Street Symposium, Anaheim AUSTROADS 1993. Guide to t raffic e ngineering p ractice, Part 6 -Roundabouts. Sydney. Courage, K. 1997. Roundabout m odeling in CORSI M In : Proceedings of the Third International Symposium on Intersections without Traffic Signals, Portland Flannery, A., Elefteriadou L. Koza P. and McFadden J. 1998. Safety, delay and c apacity of s ingle l ane roundabouts in the United States. Transportation Research Record, Journal of the Tra nsportation Research Record 1646, 63 70. Guichet, B. 1997. Roundabouts in France: d evelopment, s afety, d esign and c apacity In: Proceedings of the Third International Symposium on Intersections w ithout Traffic Signals, Portland. Highway Capacity Manual 2000. T ransportation Research Board, Washington, DC, 2000. Holm, P., Tomich, D., Sloboden, J., and Lowrance, C. 2007. Traffic Analysis Toolbox Volume IV: Guidelines for a pplying CORSIM m icro -simulation m odeling s oftware. Report FHWA HOP 07079, Federal H ighway Administration Washington DC Jacquemart, G. 1998. Synthesis of h ighway p ractice 264: m odern roundabout p ractice in the United States. National Cooperative Highway Research Program Washington DC Joseph, S. 1996. A p rocedure for c omparative e v aluation of i ntersection c ontrol m odes Master of Engineering Paper, University of Florida. Kinzel, C. ; Trueblood, M ., 2004. The e ffects of o perational p arameters in the s imulation of roundabouts In: Proceedings from the Transportation Research Board Annual Meeting Washington DC. Kimber, R. M. 1980. The t raffic c apacity of roundabouts TRRL Laboratory Report 942, Transport and Road Research Laboratory, Crowthorn e. Mini tab, 2008. Minitab Version 15. Minitab Inc., State College. Oketch T., Delsey M., a nd Robertson D. 2004. Evaluation o f p erformance o f m odern roundabouts u sing P ARAMICS m icro -s imulation m odel In: Proceedings of the TAC Annual Conference, Quebec City.

PAGE 81

81 Persaud, B., Retting, R., Garder, P., and Lord, D., 2000. Crash reductions f ollowing i nstallation of r oundabouts in the United States Insurance Inst itute for Highway Safety Report Robinson, B. W., Rodegerdts L. Scarbrough W. Kittelson W. Troutbeck, R. Brilon W. Bondzio L. Courage K. Kyte M. Mason J. Flannery, A. Myers E. Bunker J. and Jacquemart G. 2000. Roundabouts: a n i nformational g uide Report FHWA -RD 00 067, Federal Highway Administration, Washington DC Rodegerdts, L., Blogg, M., Wemple, E., Myers, E., Kyte, M., Dixon, M., List, G., Flannery, A., Troutbeck, R., Brilon, W., Wu, N., Persaud, B., Lyon, C., Harkey, D. and Carter, D., 2007, NCHRP 572: Roundabouts in the United States National Cooperative Highway Research Program Washington DC. Stanek, D ., Milam, R., 2005. High -c apacity roundabout i ntersection a nalysis: g oing s round in c ircles. In: Proceedings of the National Roundabout Conference Vail Todd, K., 1991. A h istory of r oundabouts in Britain T ransportation Quarterly, Vol. 45, No.1, 143155. Trueblood M. and Dale J., 2003. Sim ulating roundabouts with VISSIM. In: Proceedings of the 2nd Urban Street Symposium Anaheim. University of Florida, 2007. Data Acquisition for Intersection and Traffic Applications (DAITA) Version 2.0. McTrans, Gainesville. University of Florida, 2008. Traffic Software Integrated System (TSIS -CORSIM) Version 6.1. McTrans, Gainesville.

PAGE 82

82 BIOGRAPHICAL SKETCH Aaron Elias grew up in Orlando, FL where he graduated from William R Boone High school in 2002. He began h is post -secondary education at the University of Florida in the fall of 2002 graduating Cum Laude with a Bachelor of Science in civil e ngineering in the fall of 2007. In January 2008, he beg an work on his m asters d egree in t ransportation e ngineering at the University of Florida and graduated in August 2009