1 TWO LANE HIGHWAY SIMULATION AND ANALYSIS By JING LI A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2012
2 201 2 Jing Li
3 This work is dedicated to my p arents and friends
4 ACKNOWLEDGMENTS Pursuing higher education in the US wa s the biggest decision I ha ve ever made and this decision was driven by my dreams. I know that t he journey to wards realizing dreams is not an easy one Therefore, I a m so grateful and thankful at this moment for being able to see my dream come true. I will never forget the people who helped me during this journey. Without your support, caring, guidance and encouragement, I would never have gotten this far. First of all, I would like to express my appreciation to my adviser Dr. Scott Washburn, for his kindness, continuing support, and inspir ing and invaluable guidance I would also like to thank Dr. Lily Elefteriadou Dr. Yafeng Yin, Dr. Sivaramakrishnan Srinivasan Mr. William Sampson and Dr. Ruth Steiner for serving as my dissertation committee members. Their k ind assistance and thoughtful and valuable comments during my time as a doctorate student have provided me with great support. Many thanks also go to Mr. Tom Simmerman, Dr. David Hale, and Mr. Bin Lu for their technical support which was of considerable v alue to my research. I would like to say thank you again to my parents for giving me endless love and support as well as for always encouraging me to pursue my dreams. grateful to my friends in China, especially Jeff Shangguan, Fei Song, C huanbo Dong, and Shuying Gao, for their selfless help and support, no matter how far apart we are. And I would like to thank my friends in Gainesville for making me feel at home here I am a lso honored that I can work with many wonderful fellow students at the UF Transportation Research Center and be their friends. This journey was not easy, so thank you all for giving me such a memorable and splendid experience.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURE S ................................ ................................ ................................ ........ 10 LIST OF ABBREVIATIONS ................................ ................................ ........................... 12 ABSTRACT ................................ ................................ ................................ ................... 14 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 16 1.1 Ba ckground ................................ ................................ ................................ ....... 16 1.2 Problem Statement ................................ ................................ ........................... 16 1.3 Objective and Tasks ................................ ................................ ......................... 17 1.4 Chapter Organization ................................ ................................ ........................ 19 2 LITERATURE REVIEW ................................ ................................ .......................... 20 2.1 Two Lane Highway Simulation Tools ................................ ................................ 20 2.1.1 Simulation of Two way Traffic on Isolated Two Lane Road (1967) ......... 20 2.1.2 TRARR (1980s) ................................ ................................ ....................... 21 2.1.3 TW OPAS (1998) ................................ ................................ ..................... 22 2.1.4 SLX Based Two Lane Road Simulation Model (2001) ............................ 23 2.1.5 TWOSIM (2006) ................................ ................................ ...................... 2 5 2.1.6 Simulation of Mixed Traffic Flow on Two Lane Road (2008) ................... 26 2.2 Analytical Methodologies for Analyzing Two Lane Highway Operations ........... 27 2.2.1 HCM Methodology and Related Studies ................................ .................. 27 126.96.36.199 HCM methodology ................................ ................................ ......... 27 2.2. 1.2 NCHRP Project 20 7 (160) ................................ ............................. 30 188.8.131.52 New approaches to estimating percent time spent following ......... 31 2.2.2 Percent Delay Based Methodology ................................ ......................... 34 2.2.3 South Africa Methodology ................................ ................................ ....... 36 2.2.4 Threshold S peed Based LOS Criteria ................................ ..................... 37 2.2.5 Queuing Relationship Based Methodology ................................ .............. 38 2.2.6 Empirical Study Based Performance Measures ................................ ...... 42 184.108.40.206 The Montana studies ................................ ................................ ...... 42 220.127.116.11 The Oregon study ................................ ................................ .......... 44 2.2.7 Driver Comfort Service Measures ................................ ............................ 45 2.3 Traffic Models of Two Lane Highway Operations ................................ ............. 46 2.3.1 Passing Sight Distance Model ................................ ................................ 46
6 2.3.2 Desire to Pass Model and Gap Acceptance Model ................................ 50 2.3.3 Desired Speed ................................ ................................ ......................... 52 3 DEVELOMENT OF TWO LANE HIGHWAY SIMULATION TOOL .......................... 58 3.1 Develop Traffic Flow Modeling Routines for Two Lane Highway Operations ... 58 3.1.1 Car Following ................................ ................................ .......................... 58 3.1.2 Pa ssing in an Oncoming Lane ................................ ................................ 58 18.104.22.168 When will a vehicle attempt to pass a vehicle in front of it? ........... 59 22.214.171.124 Constraints governing whether a pass will be initiated ................... 62 126.96.36.199 How a potential passer executes its passing maneuver? ............... 67 3.1.3 Passing in a Passing lane Section ................................ .......................... 72 3.2 Integrating Signalized Intersections into a Two Lane Highway Facility ............. 73 3.3 Inputs and Outputs of the Simulation Program ................................ ................. 75 3.3.1 New Inputs ................................ ................................ .............................. 75 3.3.2. New O utputs ................................ ................................ ........................... 75 188.8.131.52 Performance measure outputs ................................ ....................... 75 184.108.40.206 Passing statistics outputs ................................ ............................... 76 3.4 Preliminary Tests on t he New Simulation Program ................................ ........... 77 3. 4 .1 Experimental Design ................................ ................................ ............... 77 3. 4 .2 Findings of Data Analysis ................................ ................................ ........ 77 220.127.116.11 Basic segment tests ................................ ................................ ....... 78 18.104.22.168 Two lane highway with a passing lane tests ................................ .. 81 22.214.171.124 Two lane highway segment with a signalized intersection test ...... 82 3. 4 .3 Conclusions of the Preliminary Tests ................................ ...................... 83 4 DEVELOPMENT OF AN ANALYTICAL METHODOLOGY FOR TWO LANE HIGHWAY FACILITY ANALYSIS ................................ ................................ ......... 108 4.1 Outline of Methodology ................................ ................................ ................... 108 4.2 Simulation Experiments ................................ ................................ .................. 109 4.2.1 Testing Facility ................................ ................................ ...................... 110 4.2.2 Simulation Scenarios ................................ ................................ ............. 111 4.3 Vehicle Trajectory Based Data Processing Procedure ................................ ... 114 4.3.1 Determine Upstream Effective Length ................................ ................... 115 4.3.2 Determine Downstream Effective Length ................................ .............. 116 4.4 Model Development ................................ ................................ ........................ 118 4.4.1 Upstream Effective Length Model Development ................................ ... 118 4.4.2 Downstream Effective Length Model Development ............................... 122 4.5 Examination of the Efficacy of Percent Delay as an Appropriate Service Measure for Two Lane Highway Facilities ................................ ......................... 124 4.6 Overall Evaluation Methodology ................................ ................................ ..... 129 4.6.1 Evaluation Methodology ................................ ................................ ........ 129 4.6.2 Validation of the Methodology ................................ ............................... 130 4.6.3 Percent delay based LOS criteria ................................ ......................... 132 4.7 Additional Guidance for Facility Segmentation ................................ ............... 133 4.7.1 Segmentation Guidance ................................ ................................ ........ 133
7 4.7.2 Example ................................ ................................ ................................ 134 4.8 Comparison of the New Models to the Previous Models ................................ 138 5 SUMMARY AND RECOMMENDATIONS ................................ ............................. 159 5.1 Summary ................................ ................................ ................................ ........ 159 5.2 Recommendations ................................ ................................ .......................... 161 LIST OF REFERENCES ................................ ................................ ............................. 163 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 167
8 LIST OF TABLES Table page 2 1 HCM LOS criteria for two lane highways (TRB, 2010) ................................ ....... 54 2 2 LOS criteria for two lane highway analysis ................................ ......................... 54 2 3 LOS thresholds of follower density ................................ ................................ ..... 54 2 4 Minimum acceptable level of service for two lane highways .............................. 54 2 5 Comparison of LOS boundaries in the HCM 2000 and the proposed approach ................................ ................................ ................................ ............ 55 2 6 Proposed new thresholds for level of service on two lane highways .................. 55 2 7 Fitted model for follower density ................................ ................................ ......... 55 2 8 Probability distributions of contributing parameters ................................ ............ 56 2 9 Main descriptive statistics of the PSD design values ................................ .......... 56 3 1 Components and values of passing sight distance (AASHTO, 2004) ................. 84 3 2 Revised values of the AASHTO PSD model variables for approximating the MUTCD PSD values ................................ ................................ ........................... 85 3 3 New inputs for two lane highway simulation ................................ ....................... 86 3 4 New outputs for two lane highway simulation (by direction) ............................... 88 3 5 Preliminary e xperimental design (1) facility geometry ................................ ..... 90 3 6 Preliminary e xperimental design (2) traffic inputs ................................ ............ 90 3 7 Two lane highway with a signalized intersection testing results ......................... 90 4 1 CORSIM truck types ................................ ................................ ......................... 142 4 2 Experimental design for the scenarios w/ a left turn bay ................................ .. 142 4 3 Experimental design for the scenarios w/o a left turn bay ................................ 142 4 4 Signal timing plans ................................ ................................ ........................... 142 4 5 Statistical results of the upstream effective length model w/ a left turn bay ...... 143 4 6 Statistical results of the upstream effective length model w/o a left turn bay .... 143
9 4 7 Statistical results of the downstream effective length model ............................ 143 4 8 Comparison results for validation ................................ ................................ ..... 144 4 9 LOS criteria for two lane highway facilities based on PD ................................ 144
10 LIST OF FIGURES Figure page 2 1 Desired speed distribution during holidays and weekdays (PC_PC, southbound, no rain) ................................ ................................ ........................... 57 3 1 An example of Desire to Pass for a specific driver ................................ ............. 91 3 2 Elements of passing sight distance for two lane highways (AASHTO, 2004) ..... 91 3 3 General passing maneuver process ................................ ................................ ... 92 3 4 Illustration of DNTCP and DTC ................................ ................................ ........... 93 3 5 Example of the interface between FRESIM links and the NETSIM links ............ 93 3 6 Configurations of testing facilities ................................ ................................ ....... 94 3 7 Directional average speed results for ba sic segment tests ................................ 96 3 8 Directional PTSF results for basic segment tests ................................ ............... 98 3 9 Directional follower density results for basic segment tests .............................. 100 3 10 Directional average speed results for tests with a passing lane ....................... 102 3 11 Directional PTSF results for tests with a passing lane ................................ ...... 104 3 12 Directional follower density results for tests with a passing lane ...................... 106 4 1 Trajectories of intersection affected vehicles ................................ .................... 145 4 2 Residual histogram of the upstream effective length model w/ a left turn bay .. 146 4 3 Residual histogram of the upstream effective length model w/o a left turn bay 146 4 4 Comparison of predicted upstream effective lengths using two estimation models ................................ ................................ ................................ .............. 147 4 5 Residual histogram of the downstream effective length model ......................... 1 47 4 6 Directional speed flow results ................................ ................................ ........... 148 4 7 Directional PTSF flow results ................................ ................................ ........... 148 4 8 Directional PD flow results ................................ ................................ ................ 149 4 9 Directional follower density flow results ................................ ............................ 149
11 4 10 Directional PFFS flow results ................................ ................................ ........... 150 4 11 Trajectory of delayed vehicle on time space plane ................................ ........... 151 4 12 Illustration of a two lane highway facility with two signalized intersections ....... 153 4 13 Comparison of CORSIM ATS and HCM ATS ................................ ................... 153 4 14 PD LOS criteria based on PTSF LOS criteria for Class I two lane highways ... 154 4 15 PD LOS criteria based on ATS LOS criteria for Class I two lane highways ...... 155 4 16 PD LOS criteria based on PTSF LOS criteria for Class II two lane highways .. 156 4 17 PD LOS criteria based on PFFS LOS criteria for Class III two lane highways 157 4 18 Illustration of facility segmentation ................................ ................................ .... 158
12 LIST OF ABBREVIATION S AASHTO American Association of State Highway and Transportation Official (represents the book A policy on Geometric Design of Highways and Streets by AASHTO in this document ) APL Average platoon length ATS Average travel speed DT D riv er type (defined by CORSIM) DNTCP D istance needed to complete the pass DTC Distance to collision (with the oncoming vehicle) DTP D esire to pass EB Eastbound EPF Entering percent following FFS Free flow speed HCM Highway Capacity Manual HCS Highway Capacity Software LOS Level of service MUTCD Manual on Uniform Traffic Control Devices for Streets and Highways PD Percent delay PFFS P ercent of free flow speed PHF Peak hour factor PI P ercent impeded PSD Passing sight distance PTSF Percent time spent following RT Record type (utilized in CORSIM) TNTCP T i me needed to complete passing
13 TSIS Traffic Software Integrated System (an integrated development environment that contains CORSIM) TSW TP Time spent wanting to pass WB Westbound WTMO W illingness to move over
14 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy TWO LANE HIGHWAY SIMULATION AND ANALYSIS By Jing Li August 2012 Chair: Scott S. Washburn Major: Civil Engineering With the rapid growth of urban areas and the corresponding traffic demands, rural areas are experiencing significant growth as well. This growth is now resulting in congestion on facilities that did not have any before. One area that is becoming a concern, particularly in Florida is rural areas transitioning into more developed area s Access to these areas is usually by two lane highways and sometimes there may be an occasional traffic signal within such facilities. Appropriate analysis tools are very important for researchers to evaluate the operations on complex two lane highways However, t he current state of the art two lane highway simulation tool TWOPAS is not capable of modeling signalized intersections within a two la ne highway facility. Meanwhile the two lane highway analysis procedure in the current Highway Capacity Manual (HCM) does not account for interruptions to the flow, such as from signals. To further investigate the traffic operation within these areas, a simulation tool and a facility level analysis methodol ogy are desired In this study two lane highway modeling capability i s incorporated into CORSIM by making the necessary additions and modifications to the existing CORSIM software
15 code base. The primary issues addressed here are modeling passing maneuvers in an oncoming lane or a passing lane section Preliminary testing in this part showed that the results produced by the developed two lan e highway modeling logic are reasonably consistent with expected traffic flow theory and fi eld observations discussed in the literature. Based on the new modeling capability, an analytical methodology for analyzing two lane highway f acilities with various features was developed based upon the previous study done by Yu and Washburn (2009) Specif ically, the ve hicular data generated by CORSIM was used to distinguish the influence area of a signalized intersection from basic two lane highway segments to accomplish facility segmentation. A two lane highway facility combined with signalized intersecti ons can then be evaluated based on the aggregate service measure(s) over different types of segments. The service measure percent delay was examined and verified for its efficacy in assessing the operations on complex two lane highway facilities, especiall y the one s with signalized intersections.
16 CHAPTER 1 INTRODUCTION 1.1 Background With the rapid growth in urban areas and the corresponding traffic demands, rural areas are experiencing significant growth as well. This growth is now resulting in congestion on facilities that did not have any before. One area that is becoming a concern, particularly in Florida, is rural areas transitioning into more developed area s Access to these areas is usually by two lane highways. In some cases, there may be an occasional traffic signal and possibly segments of multilane highway along such faciliti es. Managing growing traffic demands is always a task for transportation planners and engineers and appropriate analysis tools are very important for researchers to do such analysis. However, no analysis tool exists at present for analyzing two lane highw ay facilities with occasional intersections. The signal spacing and other general characteristics of these roadways do not fit with the analysis criteria for signalized arterials in the current Highway Capacity Manual ( TRB, 2010 ), and the two lane highway analysis procedure does not account for interruptions to the flow, such as from signals. 1.2 Problem Statement To further investigate the traffic operation within these areas, a simulation tool is required at first, and then a facility level analysis methodology can be developed. The current state of the art tool in two lane highway simulation is a software program named TWOPAS. This pr ogram performs a microscopic and stochastic simulation, and provides the ability to include considerable detail about two lane
17 segments in the modeling process. However, TWOPAS is not capable of modeling signalized intersections within the modeled two lane highway facility. Meanwhile, several other simulation programs capable of simulating two lane roads with signalized intersections (e.g., CORSIM), cannot simulate passing by utilizing the lane in the opposing direction, which is the most distinctive and si gnificant traffic operation feature of two lane highways. Therefore, a simulation tool that is capable of modeling the combination of two lane highway segments and signalized intersections is desired. Afterwards, an analytical methodology for analyzing two lane highway facilities with various features particularly the combination of two lane highway segments and signalized intersections can be developed based on the simulation program developed in the first stage. The new methodology uses the previous stu dy done by Yu and Washburn (2009) as the starting point. As the previous methodology was developed based on the hybrid simulation approach it is expected to be validated and updated based on the new simulation program with the capability of modeling both two lane highway segments and signalized intersections. 1.3 Objective and Tasks The objective of this study is to develop a simulation tool that is capable of modeling two lane highway segments with various features, and then based on it, to develop an an alytical methodology for analyzing complex two lane highway facilities, particularly the combination of two lane highway segments and signalized intersections. The following tasks are carried out to accomplish the objective of this study. Task 1: Develop t he logic models for simulating passing maneuvers on two lane highways, including vehicle passing in opposing lane and vehicle passing in passing lane.
18 Task 2: Develop mechanisms for combining two lane highway segments and various features, such as signaliz ed intersections, passing lanes, etc. Task 3: Add new inputs and outputs to CORSIM for allowing users to define inputs and obtain outputs as desired. Task 4: Review the methodology developed by Yu and Washburn (2009) and set up experimental designs for analyzing the effect of a signalized intersection on a two lane highway facility. Task 5: Run the simulation experiments set up in Task 4, and prepare the simulation data for model development. Task 6: Derive the models for facility segmentation that is pa rt of the new methodology by using a statistical approach on the basis of the simulation data obtained in Task 5. Task 7 : Examine the efficacy of percent delay as an appropriate service measure for t wo lane h ighway f acilities Task 8 : Develop the procedure of the analytical methodology and validate the methodology. With this study done, a simulation tool that is capable of modeling the combination of two lane highway segments and signalized intersections becomes available. And together with the new analytic al methodology, traffic operations on complex two lane highways (e.g., two lane highway with occasional signalized intersections) can be analyzed. As the new capability is incorporated in CORSIM, it will allow users to build comprehensive traffic networks including a variety of facility types for different purposes of analysis.
19 1.4 Chapter Organization This document consists of four chapters beyond the Introduction Chapter. Chapter 2 gives an overview of the previous studies on two lane highway simulation tools, analytical methodology for evaluating two lane highway operations, and traffic models of two lane highway operations. Chapter 3 describes t he development approach for the new two lane highway simulation tool in detail, including the logic and models employed in CORSIM to facilitate two lane highway modeling. Chapter 4 is devoted to the development of the analytical methodology for analyzing t wo lane highway facilities including signalized intersections Finally, the conclusions are summarized and the future study is proposed in Chapter 5
20 CHAPTER 2 LITERATURE REVIEW This chapter summarizes the current achievements in two lane highway s imulation development, analytical methodology for evaluating two lane highway operations, traffic models of two lane highway operations, and field data collection. The overview of the literature in each aspect is given in the following sections 2.1 Two La ne Highway Simulation Tools This section summarizes the development of two lane highway simulation tools since the 1960s. The material is organized chronologically in order to facilitate a better understanding of the development of the two lane highwa y simulation tool in the past few decades. 2.1.1 S imulation of Two way Traffic on Isolated Two Lane Road (196 7 ) Paul Warnshuis ( 196 7 ) constructed a computer simulation that model ed each individual vehicle s behavior directly This simulation was intended to serve as a tool helping develop a theoretical description of the interaction between the two lanes and how th at interaction influences the traffic flow in each lane. The simulation was coded in Fortran IV and it assumed that the two lane road extends infinitely in both directions by using a two lane circular track and does not have any restrictions on speed and passing. The inputs, the flow rate in each lane, the distribution of the desired speed, the initial ordering of vehicles and the initial spacing of vehicles could be specified by users. Each vehicle, based on other assumptions, travels at a fixed desired speed except for the following or passing condition. Passing maneuvers in the simulation were governed by the rules specified in the paper A car that intends to pass another car may do so only if its leader has a relatively
21 lower desired speed, the oncoming vehicle is far enough for the vehicle to complete pass and the gap in front of the passed vehicle is sufficient for the vehicle to return to the normal lane after passing. Several other constraints were also made, such as the determination of the safe distance ahead of the passed vehicle and to the first oncoming vehicle when the passing is completed, that a vehicle may pas s only one vehicle at a time and that a vehicle may be passed by only one vehicle at a time. The findings from the simulation were that : 1) changes in the distribution of the desired speed seem ed to not affect the relationship between average speed and de sire d speed 2) the average speed decreased towards the minimum desired speed when the density increased and 3) the number of platoons did not increase significantly as the density increase d However, the logic of the passing maneuver s described by Wa rnsh uis s eems too simple to replicate reality because it only considers three factors when deciding whether a vehicle initiates passing or not. It also relied on several other unrealistic assumptions such as the re being the same density for both direction s that the vehicles have a uniform length that vehicle s can only a ccelerate and decelerate at 4 mi/h/s etc These assumptions also narrow ed the scope of application. 2.1.2 TRARR (1980s) TRARR (TRAffic on Rural Roads) was developed in the 1970s and 1980s b y the Australian Road Research Board and designed for two lane rural highways. TRARR is also a microscopic simulation model tracking each generated vehicle as it travels through the whole segment. The interactions between a vehicle and road conditions or o
22 provides more user definable variables in order to allow one to create scenarios with different road and traffic conditions. Koorey ( 2002 ) stated that TRARR uses four main input files to describe th e situation to be simulated: ROAD is used to define a two lane highway section with desired road geometry, auxiliary (passing) lanes, and no passing lines; TRAF includes the information on traffic volume and composition, simu lati on time and vehicle speeds; VEHS contains operating characteristics of vehicle fleets, such as engine power, mass and fuel consumption; OBS functions as a detector input file, which specifies the locations along the highway where data is to be recorded. T he outputs produced by TRARR are summarized in the output file OUT. The major outputs reported are mean speed, travel time, frustration (via time spent following) and VOC (Vehicle Operating Costs) benefits result ing from passing. Koorey also pointed out t hat TRARR can also be used to evaluate the benefits of adding passing lanes depending on the various spacing between passing lanes. As the simulation with more details and flexibility than its predecessors TRARR lacks the ability to handle varying traffic flows down the highway yet, particularly due to major side roads or signalized intersections. 2.1.3 TWOPAS (1998) The latest version of TWOPAS (TWO lane PASsing) was revised by Leiman et al. in 1998, as the basis for the analysis procedure modifications in the Highway Capacity Manual 2000 ( TRB 2000). Currently, the TWOPAS simulation model has been built in to the Traffic Analysis Module (TAM) of the Interactive Highway Safety Design Model (IHSDM) (ITT Corporation 2007) As a microscopic simulation model for two lane, two way highways, the position of each vehicle simulated in TWOPAS is reviewed every
23 second. The manner of each vehicle is determined by the models replicating the real world. Unlike TRARR, TWOPAS ca n allow users to specify up to 13 vehicle types, percent of traffic platooned and distribution of desired speeds for unimpeded driver speed choices. Field data were used to model 1) variations in driver performance and preferences; 2) decisions of initiat ing, continuing or aborting passing maneuvers, and decision of returning to normal lane s ; 3) lane choice, lane changing and passing in a passing lane section It should be noted that the IHSDM documentation states that by experience a directional flow rate of greater than 1700 veh/h usually causes congestion that stop s the program (ITT Corporation 2007) As for output a very detailed output file will be generated when simulation runs complete It includes simulated flow rate, percent time spent following, average travel speed, trip time, traffic delay, geometric delay, total delay, number of passes, vehicle kilometers traveled and total travel time. Similar to TRARR, TWOPAS does not currently simulate traffic turning on or off the highway at intersections and driveways. And since TWOPAS was developed with U.S. data, it is considered to be generally more applicable for modeling conditions in the U.S. than TRARR. 2.1.4 SLX Based Two Lane Road Simulation Model (2001) Lemessi (2001) developed a car following a nd lane changing microscopic simulation model of a two lane road section in SLX (Simulation Language with eXtensibility) to qualify and quantify the environmental impact on traffic. Based on the car following model in this paper, a driver tries to maintain his desired speed if the distance to its leader is greater than the driver specific critical
24 value If this distance headway becomes less than the critical value, the driver decides whether to pass the slower leader or to decelerate and follow the slower leader at the lower speed. Lemessi aggressiveness, which is his desired speed. Desired speed is link dependent and assigned to each vehicle individually, following a user definable normal distri bution over the link average free flow speed. User s can also define the parameters related to vehicle type and lane and flow regime that determine the uniform distribution used to generate entry vehicles. Another model developed in this study is the lane changing model, which considers both deterministic and stochast ic psycho physical aspects. It is stated that a vehicle can be in three lane changing scenarios: 1) the driver perceives a slower vehicle ahead in its free flow condition and chang es lane without decelerating; 2) the driver stops the conscious braking and changes lane at the speed less than his desired speed; 3) the driver changes lane at an approximately same speed as its leader when in the following mode. An overtaking factor is a ssigned to each vehicle for comparing with a given threshold value to prevent the over frequent lane changes. The threshold values can be specified for different vehicle types. If the overtaking factor is greater than the threshold value, several checks wi ll be made to determine whether the driver can improve the present situation by changing lane or not. The simulation model does not allow the same vehicle to execute two consecutive passing maneuvers less than 15 seconds apart Two output files are provide d by this simulation model. One contains records of
25 information on rand om variables, which include running time, lost time, macroscopic flow, macroscopic density, a nd macrosc opic speed. reliably creates traffic operations on two lane highways in the simulation world, but it is not comprehensive enough to simulate the various features that are typical to two lane highways, such as passing lanes and isolated inter sections. 2.1.5 TWOSIM (2006) Kim (2006) indicated that it is difficult to determine capacity for a two lane highway from field data because finding two lane highway sites operati ng at or near capacity is not easy Therefore, he developed a new microscopic simulator with MATLAB 6.5, called TWOSIM. T WOSIM is intended to be used for estimating the capacity for two lane, two way highways under a variety of prevailing geometric and traffic conditions. Fi eld data, such as vehicle count speed, vehicle length, he adways, directional distribution, lane width, shoulder width, and terrain type were collected for the calibration of input parameters in the newly developed simulation model. TWOSIM was developed in three stages : 1) TWOSIM I with a straight tangent level s egment, no opposing traffic, no additional traffic entering or exiting within a given segment, and no passing maneuvers ; 2) TWOSIM II with an additional algorithm for overtaking behavior ; 3) TWOSIM III with more advanced options that consider the presence of a driveway, horizontal curves, grades along a given segment, and truck traffic. The logic for the primary operations on two lane highways (i.e., car following, passing) is following model, AASHTO (A policy on Geometric Design of Highways and Streets) criteria of minimum passing sight distance, and time hypothesis as to the stimulus perceived by drivers in making stream crossing gap acceptance decisions.
26 : 1) the capac ity of two lane, two way highways is found to be a function of average free flow speed; 2) the presence of a passing zone does not affect capacity; 3) the directional capacity is independent of the opposing flow rate; and 4) the capacity varies due to the variations in the number of trucks and driveways, and changes in horizontal curvature and grade. Therefore, he suggested that the HCM should present varying directional capacities as a function of road and traffic conditions. He also suggested considering the influence of large trucks following model 2.1.6 Simulation of Mixed Traffic Flow on Two Lane Road (2008) Based upon the field data collected at different locations of two lane road s in India, Dey et al. ( 2008 ) developed a simulation program in the Visual Basic language. The authors indicated that this simulation can be used to determine the capacity of a two lane road and to study the effect of traffic mix on capacity and speed. Also, the effect of directional split on capacity and the effect of slow moving vehicles on capacity can be estimated by using this simulation. For the inputs, Poisson distribution was used to generate the arrival pattern according to the data collected and the normal distribution for th e basic desired speed (the maximum speed the specific vehicle can reach) assignment to each vehicle. Also, the average rate of acceleration during a passing was found to be exponential function of passing time from the field data. Additionally, t he simulat ion allows users to modify the percent compositions of the six categories of vehicles. The passing logic in this simulation program is similar to others, but it was presented in t he sequential states that describe a passing maneuver executed by a pass ing v ehicle from beginning to end. For the practical reasons, t he maximum number
27 of vehicles that can be passed by a vehicle at a time is 3 At the same time, an initiated passing can be rejected by the driver as well in the simulation. However, the critical po int beyond which a passing vehicle cannot reject passing was not presented in the paper. For the vehicle traveling at its free flow speed, the simulation will evaluate the possibi lity of flying passing when it is catching up with a slow vehicle. Although t he critical position where the vehicle has to decide whether to execute flying passing or to decelerate and follow the leader was discussed in the paper, the details of evaluating the possibility of flying passing was not provided. 2.2 Analytical Methodolo gies for Analyzing Two Lane Highway Operations This section provides an overview of the existing methodologies for analyzing traffic operations on two lane highways, as well as other performance measures proposed in the previous studies for assessing level of service on two lane highways. 2.2.1 HCM Methodology and Related Studies 126.96.36.199 HCM methodology Chapter 15 of the Highway Capacity Manual (HCM) (TRB 2010) provides an operations analysis methodology which allows for a single direction of traffic flow for two lane highway segments. The analysis procedures can also be applied to segments with the general terrain classification of level or rolling. It is stated that the directional methodology must be applied to segments in mountainous terrain or with grades of 3 % or more with a length of 0.6 miles or more, and/or segments with a passing lane. The performance measures of Average Travel Speed ( ATS ) and Percent Time Spent Following ( PTSF ) are used to determine the level of service (LOS) for Class I two la ne highways, on which drivers care about mobility the most O nly PTSF is employed to determine the LOS for Class II two lane highways with drivers less attention to efficient
28 mobility. For the newly added Class III two lane highways, which represent roadw ays in rural developed areas or scenic roadways, percent of free flow speed ( PFFS ) is the primary service measure The LOS criteria for two lane highways are presented in Table 2 1. LOS F indicates the situation when the demand exceeds the capacity. The ca lculations for ATS and PTSF for directional two lane highway segments without a passing lane are given by Eq uation s 2 1 to 2 4 ( 2 1 ) ( 2 2 ) ( 2 3 ) ( 2 4 ) where ATS d = average travel speed in the analysis direction (mi/h) FFS d = free flow speed in the analysis direction (mi/h) v d = passenger car equivalent flow rate for the peak 15 minute period i n the analysis direction (pc/h) v o = passenger car equivalent flow rate for the peak 15 minute period i n the opposing direction (pc/h) f np = adjustment for percentage of no passing zones in the analysis direction B PTSF d = base percent time spent foll owing in the direction analyzed PTSF d = percent time spent following in the direction analyzed PFFS d = percent of free flow speed in the direction analyzed
29 For segments that contain passing lanes, the ATS calculated and PTSF values are further adjusted by Eq uation s 2 5 and 2 6 ( 2 5 ) ( 2 6 ) where ATS pl = average travel speed for the entire segment in cluding the passing lane (mi/h) PTSF pl = percent time spent following for the entire seg ment including the passing lane L t = total length of analysis segment (mi) L u = length of two lane highway up stream of the passing lane (mi) L d = length of two lane highway downstream of the passing lane and b eyond its effective length (mi) L pl = length of the pas sing lane including tapers (mi) L de = length of the passing lane f pl = factor for the effect of a passi ng lane on average travel speed However, the analysis methodologies in Chapter 15 of the HCM 20 1 0 are not suitable for analyzing situations where two la ne highways are combined with o ccasional signalized intersections. I t is only suggested that isolated signalized intersections on two lane highways can be evaluated with the methodology in Chapter 18 of the HCM
30 20 1 0 I f t wo lane highways are located in urban and suburban areas with multiple signalized intersections at spacing of 2.0 mi or less the methodologies of Chapter 16 of the HCM 20 1 0 17 of the HCM 20 1 0 will be applied. 188.8.131.52 NCHRP Project 20 7 (160) Harwood, et al. ( 2003 ) noticed that the users encountered two issues regarding the two lane highway operational analysis procedure in the HCM 2000 This resulted in a study done by Harwood and his associates to address these issues. The first issue is that users found lower service volumes for two lane highway segments in the HCM 2000 than in the HCM 1997. Harwood, et al. determined that overestimating percent time spent following ( PTSF ) by the HCM 2000 methodologies resulted in this issu e, which was proved by the mathematical comparisons. Based on the cause of the issue, they suggested revising the adjustment to directional PTSF for percentage of no passing zones as follows : ( 2 7) where f adj is the adjustment factor for directional split and no passing zones As a result, the equation for estimating directional PTSF should be revised as: ( 2 8) With this solution, the undesirable difference between the two way and directional segment procedures can be eliminate d. The second issue identified by Harwood, et al. is that users found that the classes of two lane highways defined in the HCM procedures are not applicable to two lane
31 highways with various features, such as a reduced speed limi t when approaching urban or suburban area s and scattered signals on the urban extension of a two lane highway. The researchers believed that such areas of interest ( termed as developed areas in the report ) are difficult to refer to any class of two lane highways due to complex driver expectations. The field observations collected on the two lane highways in Florida verified their opinions. To improve the HCM procedures, the researchers suggest ed that two lane highways in developed areas can be defined as a new type for the purpose of analysis either appearing as a new class in Chapter 20 of the HCM 2000 or in a new chapter. Another proposed solution is to expand Chapter 15, Urban Streets of the HCM 2000 so that it contains an analysis procedure for the se developed arterials The details of each approach discussed in the report are briefly summarized as follows : Incorporate the concepts from Chapter 15 of the HCM 2000 into Chapter 20 to accommodate two lane highways in developed areas. The concepts from Chapter Signalized Intersections Un Signalized Intersections of HCM 2000 may also be utilized in this incorporation for analyzing situations wit h intersections. Develop a speed prediction model that is capable of estimating urban arterial ATS more precisely in order to analyze urban streets and two lane highways in developed areas. Create a new chapter that contains an assessing methodology for tw o lane highways and multilane highways in developed areas. 184.108.40.206 New approaches to estimating percent time spent following Regarding the percent time spent following overestimation problem in the HCM 2000 methodologies, Durbin (2006) proposed two new app roaches to estimating the percent time spent following ( PTSF ). I t is suggested that only vehicles that are impeded by slower leading vehicles and not able to travel at their own desired speeds or pass
32 slower vehicles contribute to PTSF Based on this idea, the author provided two methods for estimating PTSF Weighted average approach Durbin indicated that the proportion of passenger cars traveling involuntarily in platoons is virtually identical in meaning to PTSF For estimating this proportion, tw o assumptions were made. First i t was assumed that heavy vehicles never follow passenger cars, but the converse is not true. Second, it was assumed that passenger cars traveling involuntarily in platoons travel at the average speed equal to about the averag e speed of the leader. Thus, the formula was developed as follows : ( 2 9 ) where P pc2 = proportion of passenger cars involuntarily traveling in platoons S atot = average actual travel speed of all vehicles S dpc = average desired travel speed of passenger cars P pc = proportion of passenger cars in the traffic mix S ahv = average actual travel speed of heavy vehicles P hv = proportion of heavy vehicles in the traffic mix Probabilistic approach This approach attempted to estimate PTSF by using probabilities. To implement this approach, slow moving vehicles and fast moving vehicles were defined for the vehicle classification. Any vehicle regardless of its size and performance can be placed in one of the two categories. S low moving vehicles usually lead or potentially lead platoons while fast moving vehicles are usually impeded by slow moving vehicles and join platoons for lack of passing opportunities. The
33 relationship between the two vehicle categories was then used to estimate t he value of PTSF Two probabilities were included in this estimation, one of which represents the probability that a vehicle tra vels in a platoon and the other represents the probability that a platooned vehicle involuntarily travels at a speed lower than the desired speed. Therefore, PTSF can be estimated by ( 2 10 ) where P p = probability of a vehicle being part of a vehicular platoon P t = probability of a vehicle traveling involuntarily in a platoon at a speed lower than the desire d speed The details of the way to measure these two probabilities were also discussed in the paper. The probability regarding vehicles traveling in a platoon was suggested being measured by using vehicle time headway data. Vehicles with time headways less than a pre defined threshold value are considered inside platoons, either involuntarily or voluntarily. The procedure for estimating the probability that a vehicle involuntarily travels in a platoon in the field was developed based upon the assumptions tha t the distribution of desired speeds is the same for both vehicles in platoons and vehicles in free flow conditions, which is usually a normal distribution. This procedure first measures the average travel spe ed of slow vehicles, and then the distribution of desired speeds for all vehicles can be established, based on which vehicles with speeds higher than the averag e speed of slow moving vehicles can be identified. The author stated that both methods could provide reasonable PTSF values, which were lower t han the ones with the HCM 2000 procedures. However, although i t is
34 rational to distinguish involuntary following and voluntary following when estimating PTSF by the probabilistic approach, the definitio ns of these two following types need to be further inv less than its desired speed but greater than or equal to its tolerable speed may not be considered involuntary. Thus, the distribution of tolerable speed needs to be investigated as well as the distribution of desired speed. 2.2.2 Percent Delay Based Methodology Yu and Washburn (2009) developed a facility based evaluation methodology for two lane highways. This methodology allows the various features (e.g., isolated intersections, passing lanes) that are typical to an extended length of two lane highway to be analyzed as a single facility. divide s the facility into appropriate segments, and properly account s for the operational effects of traffic flow transi tioning from one segment to another. A two lane highway with an isolated signalized intersection was used as a model in this study. For a two lane highway facility with signalized intersections, the entire length of the facility is divided into three types of segments : the basic two lane highway, the signal influence area, and the affected downstream segment. Percent delay ( PD ) was selected as the common service measure for the interrupted flow facility of a two lane highway with signalized intersections. A ccording to the paper, percent delay is calculated by div iding delay by free flow travel time. An excess of average travel time over free flow travel time for drivers on a facility is defined as d elay PD is calculated by ( 2 1 1 )
35 where PD = average percent delay per vehicle for the entire facility (%) D H = average delay time per vehicle for the two lane highway segment (s/veh) D S = average delay time per vehicle for the signalized intersection influence area (s/veh) FFS H = free flow speed for the two lane highway segment (ft/s) FFS S = free flow speed for the signalized intersection influence area (ft/s) L H = length of the two lane highway segment (ft) L S = length of the signalized intersection influence area (ft) Percent erception of freedom during driv ing. A may be affected by the presence of signal controls, restrictive road conditions (e.g., no passing zones), or opposing traffic, so these effects can also be reflecte d by percent delay. The LOS criteria for two lane highway facilities, based on percent delay, are shown in Table 2 2. The authors used a simulation based approach to develop the computational methodology, due to the difficulties of measuring the field data needed. A hybrid simulation approach combining CORSIM and TWOPAS was applied. The CORSIM simulation program was used to model a two lane roadway segment that contains a signalized intersection. T he downstream two lane highway segment relative to the signa lized intersection was then simulated by TWOPAS. The regression models for estimating the length of the effective upstream/downstream segment were developed
36 based on the simulation results. Given the regression models, the process of determining the LOS of a two lane highway facility could be outlined as follows : Step 1. Compute the lengths of the effective upstream and downstream segments and then d ivide the facility into segments accordingly Step 2 Determine the free flow speed of the facility Step 3 C ompute the average travel speed on the basic two lane highway segments Step 4 Compute control delay at the signalized or unsignalized intersection i nfluence area Step 5 Compute the average travel speed on the affected downstream segment Step 6 Dete rmine the delay for each segment by calcula ting the difference between average and free flow travel time, except for the intersection influence area using the control delay as the segment delay. Step 7 Determine the percent delay by dividing the total de lay by the total free flow travel time along the whole facility, leading to the determination of the facility level LOS However, Yu and Washburn s methodology was built upon the results of the hybrid simulation approach, therefore, i t is still necessary to use field data or an integrated simulation model to verify and validate this methodology. 2.2.3 South Africa Methodology The South African National Roads Agency Limited (SANRAL) (2006) developed an analy tical method ology for determining the level of service of two lane highways in rural areas A macroscopic simulation model was developed and extensively calibrated by the field observations in this study. Although called a macroscopic simulation model, the simulation model actually used t he microscopic method to simulate free flow conditions and the macroscopic method to simulate queue formation/platooning, respectively. The simulated platooning can be further used to estimate traffic
37 performance measures, such as average travel speed, den sity, percentage followers and follower density. T h e res earchers also investigated several alternative performance measures other than average travel speed and percent time spent following which were employed in the HCM 2000 two lane highway methodology. They identified follower density as the most suitable measure, for it contains the combined meanings of percentage followers, traffic flow and average travel speed. None of these three measures can fully reflect a driver s perception of LOS when travelin g on a two lane highway. To provide criteria for determining LOS based on follower density, the field observations were analyzed in two aspects. The researchers first developed a set of threshold values for determining the follower density based LOS, and t hen estimated the minimum acceptable LOS below which the capacity expansion of a facility needs to be considered. The details of the criteria are presented in Table s 2 3 and 2 4 It should be noted that the calibration of the simulation model and the LOS c riteria were based on the field data collected in South Africa. Therefore, the characteristics of driver behaviors there, such as passing on wide shoulders or in no passing zones, should be attended to when applying the methodology developed in this study to other countries. 2.2.4 Threshold S peed B ased LOS C riteria Romana and and Prez ( 2006) pointed out that for two lane highway travelers, their desire to pass and frustration over being delayed should not be determined only by looking at headways of the difference between actual speed and desire d speed. After reviewing the existing measures of effectiveness and other possible measures, the authors proposed a measure named threshold speed to better determine the LOS of a
38 two lane highway facility. As state d in the paper, threshold speed is defined as the minimum acceptable speed for drivers that are traveling under high flow rate condition on a road with the consistent characteristics. They proposed that threshold speed should be applied along with existing performance measures (i.e., average travel speed and percent time spent following) to determin e the LOS of two lane highways in the HCM 2000. The new LOS assessing approach developed in the paper is summarized as follows : If the average travel speed on a two lane highway is higher than the threshold value, the percent time spent following will be the only necessary measure to check to ascertain the level of service. If the average travel speed is lower than the threshold value, percentage time spent follow ing becomes less important than speed when determining the LOS of two lane highway facility. With this new approach, i t is expected that various features of different highways can be considered rather than classifying highways into Class I and II in the first place. At the same time, a driver s perception of the level of service can also be reflect ed by this approach However, the authors did not provide details o f how to determine the threshold speed, and instead only suggested several ways to select threshold speed. A field study using the new approach mentioned in the paper use d 80 km/h as the threshold speed, which approximate d the users expectations obtained from surveys in Spain. The example LOS criteria with the threshold value of 80 km/h compared with the HCM 2000 criteria in the paper are presented in Table 2 5 2.2.5 Queuing Relationship Based Methodology Two problems concerning the current HCM (TRB, 2010) two lane highway procedures were pointed out by Polus and Cohen (2009) based on their field
39 observations in Israel. One is that LOS is insensitive to average travel speed, and the other is overestimation of percent time spent following At this point, the authors developed queuing models for estimating the LO S of two lane highways on the basis of Queuing Theories. Next, they estimated the parameters of the queuing models using field data collected in northern Israel. Five major parameters were included in this study : flow, average platoon length, traffic inten sity, percent time spent following and f reedom of f low As the authors stated, all of the parameters can be used alo ne or in combination with other s to determine LOS. Some primary findings are summarized as follows : Flow A linear relationship was found between one way flow and speed according to the field data. Although only one way flow was involved, they found that the speed data had included the influence of the opposing flow. Average platoon l ength The average number of headways per platoon obtained from the field data were used to calculate the average p latoon length by simply adding one to account for the lead vehicle. ( 2 1 2 ) where APL = averag e platoon length (vehicles) Q 0 = a verage number o f headways inside platoons (dimensionless) The average number of headways inside platoons Q 0 was found to be related to one way flow as follows : ( 2 1 3 ) where V is the one way volume
40 The field data shows a positive relationship between Q 0 and one way flow, and Q 0 is significantly sensitive to flow. Thus, APL was believed by the authors to be the best parameter for measuring LOS. Traffic i ntensity The M/M/1 queuing model was used to describe the formation of pl atoons on two lane highways. T he r atio between the average wait time for the first vehicle behind a lead vehicle to pass and the average inter arrival time vehicles join the back of the platoon, which was defined as traffic intensity can be derive d as ( 2 1 4 ) T he relationship between and one way flow V was also estimated by the field data ( was measured by Eq uation 2 15 ) as follows : ( 2 1 5 ) The field data show that traffic intensity levels off after t he one way volume reaches 600 veh/ h. Percent t ime s pent f ollowing ( PTSF ) T he equation for calculating PTSF was derived on the basis of Queuing Theories as follows: ( 2 1 6 ) where PTSF = percent ti me spent following (%) N 0 = average number of headways b etween platoons (dimensionless) Again, the relationship between PTSF and two way flow was derived based on the field data, thus ( 2 1 7 )
41 where V p is the two way flow in passenger cars equivalent Freedom of Flow This parameter represents the freedom a driver perceives when traveling in the traffic stream, which thereby becomes the LOS evaluation measure. It can be calculated by ( 2 1 8 ) where is the freedom of flow A regression model expres sing the relationship between and two way flow V p was shown as ( 2 1 9 ) As the authors concluded, although all five parameters can be used to estimate LOS of two lane highways as they are highly related to the flow characteristics, PTSF was considered the most suitable measure because i t is capable of evaluatin g the cost of delay in economic terms I t wa s also suggested that freedom of flow should be an additional measure for estimating LOS as it might contain more information on perception. Based on the proposed flow measures, the new thresholds for LOS on two la ne highways were developed ( Table 2 6). The PTSF values calculated by equat ions in this paper wer e much lower than the HCM PTSF values, which is consistent with t he findings in other studies. However, since the models developed in this paper were based on the field data in Israel, the local characteristics should be taken into account when applying the methodology outside the country.
42 2.2.6 Empirical Study Based Pe rformance Measures 220.127.116.11 The Montana studies I t is well known that PTSF is difficult to measure in the field ; therefore, some researchers tried to find alternative performan ce measures that are easier to collect on two lane highways. Al Kaisy and Sarah Karjala (2008) investigated six selected performance indicators : average travel speed, average travel speed of passenger cars, average travel speed as a percent of free flow speed, average travel speed of passenger cars as a percent of free flow speed of p assenger cars, percent followers, and follower density This empirical study was based on field data collected in the state of Montana at four different study sites. The relationships between performance measures and primary platooning variables such as tr affic flow heavy vehicle percentage and no passing zone percentage were examined As a result, follower density was identified as the best measure for describing the level of service of two lane highway operations, as it can reflect the traffic level tha t was indicated as a dominant factor that determines the performance of most highways. T he authors stated that percent age of followers has no strong relationship with traffic level as it depends mainly on time headway distribution. W ays to measure followe r density in the field were also provided in the paper. However, the threshold values of follower density for determining the LOS of two lane highway operations were not discussed in the paper. In 2010, Al Kaisy and Freedman did another empirical investiga tion into a new measure Percent Impeded ( PI ) which was also intended to evaluate performance of traffic operation s on two lane highways. PI was defined in the paper as a point measure indicating the percentage of vehicles that follow slower vehicles due to lack of passing opportunities on two lane highways. The methodology for deriving PI is similar to the
43 probabilistic approach for estimating PTSF (2006) thesis In order to ca pture impedan ce experienced during driving, l ocations in the immediate upstream and downstream area of a passing lane where platoons break up and form were suggested as ideal for collecting data Besides PI the performance measures of percent followers, follower density, and ratio of average travel speed to free flow speed were also examined in this study as well for reference. PI was eventually validated in this study. The empirical analysis showed that PI is more sensitive to platooning than other performance measures. In comparison PI contains more information due to relatively high correlations with other measures and platooning variables except for traffic flow. Furthermore, PI is the only measure that shows the variation trend corresponding to the one of general performance measu re s for upstream and downstream of the passing lane presented in the HCM 2000. Four multivariate regr ession models were developed to estimate the four performance measures using the data from one of the two study sites. The one for PI is given as ( 2 20 ) where X 1 = volume in the same direction of travel X 2 = volume in the opposing direction of travel X 3 = percent no passing X 4 = distance downstream of the passing lane X 5 = presence of merge effect (limited to the first station downstream of passing lane) X 6 = percentage of trucks
44 Variables X 1 and X 2 were found to be insignificant at the 95% confidence level. The LOS criteria based on PI were not discussed in the paper, and as the authors expected, further investigation into the performance measure PI for a complete analytical methodology is need ed 18.104.22.168 The Oregon study The Facility Analysis and Simulation Team at the Oregon Departmen t of Transportation (2010) established several models for predicting alternative two lane highway performance measures due to the difficulty in collecting field data of PTSF Several performance indicators and platooning variables that could describe two lane highway operations were selected. After that, field data were collected from 13 sites in Oregon State for model development, and another 4 sites in Oregon for model validation. Based on the field data, the pre dicting models were formulated and finally validated. The performance indicators included in this study wer e average travel speed, average travel speed of passenger cars, average travel speed as a percent of free flow speed, average travel speed of passenger cars as a percent of free flow speed of passenger cars percent followers, and follower density. All the indicators were directly or indirectly obtained from the field data. It should be noted that follower density in this study was measured by ( 2 2 1 ) To es tablish the models, platooning v ariables were collected as well and served as the independent variables in the models, including traffic flow, percent heavy vehicles, standard deviation of free flow speeds, percent no passing zones and terrain.
45 As a result of statist ical analysis, follower density was finally selected as the b est performance indicator. F urther, the researchers found that follower density has a wide spectrum, which makes the level o f service categories easy to determine The final mod el is given in Tab le 2 7 This promising estimate model is of great importance as only a few studies have been done on follower density, especially on the basis of field data. The conclusions of this study may be used as references when developing two lane highway simulati on models and analytical methodologies. 2.2.7 Driver Comfort Service Measures The HCM 2000 uses density as the key indicator to determining the level of service of traffic operations on both rural and urban freeways. However, Kim, Courage et al. (2003) poi nted out that although density is ideal for assessing urban freeways which usually need to accommodate high demands, it is uncertain whether density is appropriate for assessing rural highways drivers. Therefore the researchers proposed three measures which are intuitively related to driver comfort: acceleration noise, number and duration of cruise control applications and percent time spent following. In their initial study, CORSIM was used to generate the req uired data instead of collecting them in the field. The step by step data were processed to derive the propose d measures, as summarized below: Acceleration noise. This is calculated as the standard deviation of the acceleration for ea ch vehicle produced by CORSIM. Number and duration of cruise control applications. A cruise control emulator was developed and incorporated into the previous CORSIM logic. Next, the proportion of time the cruise control was used and the number of cruise control applications was recorded and reported.
46 Percent time spent following. This following scheme developed by Halati et al. (1997) was used to determine when a vehicle is identified as a follower. T he percent time spent following was then determined. T he relationships betwee n traffic volume and the proposed measures were determined by plotting the processed data. According to these relationships, the authors concluded that acceleratio n noise can be directly used to establish LOS criteria for rural highway s. The other two meas ures considered were found to require further investigations. Although the proposed measures are promising, field data are needed to confirm the conclusions in the paper as all the findings were based on the simulation data. Also, because the measures in the paper apply to all rural highways rather than only two lane rural highways, i t is necessary to consider the characteristics of two lane lane highway analysis. 2.3 Traffic Models of Two Lane Highway Opera tions Given a logical framework for operations on two lane highways, i t is very important that the appropriate traffic models are added to each step to make the framework functional. This section reviews the previous work on passing sight distance model (s) desire to pass model (s) and gap acceptance model (s) A study on desired speed distribution is introduced at the end of this section. 2.3.1 Passing Sight Distance Model According to NCHRP Report 605 (Harwood, et al. 20 08) the current passing sight dista nce (PSD) criteria for two lane highways in A P olicy on Geometric Design of Highways and Streets (AASHTO 2004 ) is for highway design, while the criteria for marking passing or no passing zones on two lane highways are included in the Manual on Uniform Tra ffic Control Devices for Streets and Highways (MUTCD) (FHWA 2003)
47 AASHTO defines the minimum PSD required as the sum of four distances traveled by a passing vehicle: 1) distance traveled during perception and reaction time, and during initial acceleration before passing vehicle moves into the left lane; 2) distance traveled in the left lane during passing; 3) distance between passing vehicle and opposi ng vehicle when passing is completed; and 4) distance traveled by an opposing vehicle for two thirds of the time the passing vehicle travels in the left lane. The criteria are essentially the same as those in the 1954 AASHTO policy. However, the PSD value at different design speed level s proposed by AASHTO based on this definition is substantially larger than the one derived by the MUTCD models. This was explained in the report as that AASHTO defines passing sight distance in a too conservative way, which c ommits a passing driver to complete the pass; while the MUTCD criteria are appropriate for a passing driver wishing to abort the pass. Glennon (1988) derived a mathematical model for estimating the critical position for a passing driver during the passing maneuver to complete or abort the pass, when he found that the AASHTO (1984) PSD model did not consider the possibility of an based on several assumptions as follows: The oppos ing vehicle travels at the design speed; The passing vehicle accelerates to the design speed at or before the critical position; The impeding vehicle travels at a constant speed design speed; and The minimum acceptable headway is one second for both cleara nce between passing and impeding vehicles and the clearance between passing and opposing vehicles at the end of the pass.
48 As a result, the critical position model and the critical passing sight distance model were obtained as follows : ( 2 2 2 ) ( 2 2 3 ) where C = critical separation between pas sing and impeding vehicles (ft) L I = length of impeding vehicle (ft) L P = length of passing vehicle (ft) m = difference in speed between passi ng and impeding vehicles (ft /s ) v = design speed, speed of passi ng and opposing vehicles (ft/s ) d = deceleration rate of passing vehicle abortion (ft/s ) S C = critical sight distance (ft) To apply these models to the design and the marking of a passing zone, the author made some ad ditional assumptions. In this case, the PSD requirements obtained become considerably less than the AASHTO PSD requirements, but very close to the MUTCD requirements (even though the MUTCD requirements were derived with a completely different set of models and criteria). By doing another analysis of the sensitivity of PSD values to vehicle length, the author concluded that the impact of truck length is not as significant as reported in previous studies. Hassan et al. (1996) also developed a revised model fo r determining the minimum t is based on the concept of critical sight distance and developed by means of kinematic equations
49 describing the interactions between the vehicles involv ed in a passing maneuver. Three main modifications are identified in the paper: Hassan assuming that a passing driver whose front bumper is abreast with the front bumper of the passed vehic le will complete the pass even if the critical position has not yet been reached. Thus, two models are derived, for both C > 0 and C 0. Hassan model uses a more conservative assumption that the passing driver will experience an additional perception reac tion time when deciding whether to abort the passing maneuver or not. The clearance distance between the passing vehicle and opposing vehicle at the end of an aborted pass is different from that in a completed pass. The authors pointed out that it should b e related to the speeds of the passing and opposing vehicle at that moment. And the clearance distance between the passing and passed vehicle, either at the end of an aborted pass or a completed pass, is related to the speed of the trailing vehicle, rather than a constant speed differential. The revised model was finally validated using a set of field data, and the results showed that the PSD requirements are longer when a passing vehicle is trying to pass a truck for speeds up to 110 km/ h. Therefore, i t is suggested that the MUTCD standards consider the effect of vehicle length to ensure safe passes. The current PSD requirements for design and marking vary for different design speeds. However, many random variables are ignored in the calculation process. Fa cing this fact, Khoury and Hobeika (2007) derived a PSD distribution that considers the contributing random variables by devising a Monte Carlo simulation model and a closed form analytical estimation model. The base model for this development is PSD model, because it provides the best description of the mechanics of a passing maneuver. The distribution of each contributing parameter was assigned according to the literature and the field data sources, as shown in Table 2 8
50 The authors used a soft ware package named Cristal Ball to perform the random sampling from the corresponding probability density functions for each parameter mentioned in Table 2 8 The best fit distribution of the resulting PSD obtained is a Gamma distribution. The main descrip tive statistics of the critical point (the position beyond which a passing vehicle is committed to complete the pass) and the PSD values for a design speed of 80 k m/h (50 m i/h ) are presented in Table 2 9 Then the PSD distribution results were verified by a parallel solution method, which was used to obtain the first and second moments of the PSD distribution. The differences of means and standard deviations between the Monte Carlo simulation and the analytical model are less than 2%. For the conclusions, t he authors indicated that in this paper, but the AASHTO PSD values are overestimated. As the authors expected, the PSD distribution obtained in this study can be used to de velop a unique microscopic simulation, which makes it feasible to quantify the trade off between performance and safety finally. 2.3.2 Desire t o Pass Model and Gap Acceptance Model fore, a However, only a few studies have been done on this topic, for i t is difficult to capture rah and Toledo acceptance decisions to complete a desired pass. It should be noted that the passing gap in this paper is defined as the time gap between the opposing v ehicle and the subject vehicle when the subject vehicle is in the following mode. This term is
51 essentially the same as passing sight distance but by time instead of distance. The development of the model is based on the data collected with a driving simula tor model contains two levels of choices. The first level choice is the decision whether or not to consider passing according to the desire to pass, and the second lev el choice is the decision whether or not to accept an available passing gap. The desire to pass is formulated as a binary choice problem: ( 2 2 4 ) where = the choice indicator variable with value 1 if the driver de sires to pass and zero otherwise = the utility to the driver from desiring to pass ( a function of explanatory variables speed differential, following distance and cumulative distance) And if a driver desires to pass, then he/she will evaluate the available passing gaps against his/her minimum acceptable gap. The driver passes the impeding vehicle only if the available passing gap is acceptable. ( 2 2 5 )
52 where = choice indicator variable with value 1 if the gap is accepted and zero otherwise = the available passing gap = the critical passing gap (a function of explanatory variables current speed, lead speed, opposing vehicle speed, road curv ature, type of front vehicle and age) Then t he probability to complete a passing maneuver is determined by the product of the probabilities of a positive decision on both choices. The results show that a ll the variables in the model are statistically signi fic ant at the 95% confidence level Although the results reported are promising, the authors indicated that t he collection of driving simulator data may result in biases in behavior. In some cases, simulator drivers may modify their behavior if they feel indifferent or tired in the course of the experiment. And the passing behavior model proposed in this paper is not able to handle the situations such as a passing driver aborts the pass or passes multiple vehicles at a time. 2.3.3 Desired Speed User dissat isfaction is usually due to the situation that a driver cannot travel at his/her desired speed when following a slower vehicle. Therefore, Catbagan and traveling below th e desired speed because i t is impeded by the slower lead vehicle. Then they found it necessary to develop desired speed distributions to calculate a
53 would be possible to better replicate following behavior on a passing restricted two lane highway in a simulation. The study took the data collected from an ordinary two lane highway (with passing restrictions) in Japan as the basis. The data were gathered for six consecuti ve months. A video survey was also conducted at the study segment to verify the detected vehicle lengths based on vehicle classifications. Two types of vehicles were identified for the study purpose, which are passenger cars (PC, 0 to 4.9 m) and heavy vehi cles (HV, 4.9 m and longer). The authors then used the modified Kaplan Meier Approach (the original method from Hoogendoorn, 2005) to estimate free speed distributions by using the field data. The primary findings are summarized as follows : Holiday travele rs usually prefer higher desired speeds than weekday travelers. The d esired speeds of passenger car drivers are not sensitive to period of time during a day. A nd heavy vehicle drivers generally prefer higher speeds during daytime. Figure 2 1 presents the a nalysis results. Truck driver s behavior is greatly affected when following another truck o n a rainy day. Gradient has some effect on the free speed distributions The author mentioned that drivers may travel at speeds lower than the actual desire d speeds because of the slope constraints. Although the estimated desired speed distribution s for various conditions were proposed in this paper, no specific distribution function was presented for estimating parameters and adjustment factors in different situation s.
54 Table 2 1 HCM LOS criteria for t wo l ane h ighways (TRB, 2010) LOS Class I Class II Class II I Percent Time Spent Following Average Travel Speed (mi/h) Percent Time Spent Following Percent of Free Flow Speed A > 55 > 91.7 B > 35 50 > 50 55 > 40 55 > 83.3 91.7 C > 50 65 > 45 50 > 55 70 > 75.0 83.3 D > 65 80 > 40 45 > 70 85 > 66.7 75.0 E > 80 > 85 66.7 Table 2 2 LOS c riteria for t wo lane h ighway a nalysis Level of service Percent Delay (%) A B > 7.5 15 C > 15 25 D > 25 35 E > 35 45 F > 45 Table 2 3 LOS thresholds of follower density LOS "Typical" follower density Range of follower densities A 1.0 0.3 1.4 B 2.0 1.3 3.3 C 4.0 3.0 6.7 D 8.0 6.3 9.5 Table 2 4 Minimum acceptable level of service for two lane highways Two lane highway class Acceptable level of service Normal day peak hour 30th Highest hour Class I Highways LOS C LOS D Class II Highways LOS D LOS D
55 Table 2 5 Comparison of LOS b oundaries in the HCM 2000 and the p roposed a pproach LOS HCM 2000 (Class I, speed in km/h) HCM 2000 ( PTSF in %) Speeds Greater than Threshold, 80 km/h Speeds Lower than Threshold, 80 km/h A > 90 PTSF 35 % DV 30 B 80 < s 90 35 < PTSF 50 30 < % DV 55 C 70 < s 80 50 < PTSF 65 55 < % DV 75 D 60 < s 70 80 < PTSF 90 75 < % DV 60 < s 80 E 40 < s 60 80 < PTSF 40 < s 60 F s 40 % DV = percentage of delayed vehicles Table 2 6 Proposed n ew t hresholds for l evel of s ervice on two l ane h ighways Level of service PTSF thresholds (%) Two way flow thresholds(pc / h) Freedom of flow thresholds (Eta) A 0 15 0 300 16.5 B 15 30 300 700 7.1 16.5 C 30 45 700 1,200 4.1 7.1 D 45 60 1 ,200 1,800 2.8 4.1 E 60 75 1,800 2,700 1.8 2.8 F 75 100 2,700 1.8 Table 2 7 Fitted model for follower density Variable Coefficient Dependent variable Follower density Independent variable Intercept 0.4823332 Traffic v olume 0.0067640 Opposing v olume 0.0006175 % Heavy v eh i cles 0.0008791 % No Passing 0.0008097 Terrain (1 for Level, 2 for Rolling) 0.2482458
56 Table 2 8 Probability distributions of contributing parameters Parameter Name Distribution type Source V Speed Truncated normal Field data M Speed differential Log normal AASHTO, MUTCD, Polus, Glennon D Deceleration rate Truncated normal Fambro L P ,L I Vehicle lengths Discrete US Census Bureau (vehicle invt.) R Perception reaction Log normal McGee, Neuman, Lerner G C Completed pass gap Uniform AASHTO, MUTCD, Polus, Glennon G A Aborted pass gap Uniform AASHTO, MUTCD, Polus, Glennon Table 2 9 Main descriptive statistics of the PSD design values Variable Critical point PSD Confidence interval (%) 95 95 99 Mean (m) 3.75 213.65 215.32 Median (m) 3.57 209.25 210.23 Standard deviation (m) 2.08 9.99 9.45
57 Figure 2 1 Desired speed distribution during holidays and weekdays (PC_PC, southbound, no rain) Source: Catbagan, Nakamura. Desired Speed Distributions on Two Lane Highways Under Various Conditions, Transportation Research Record Washington D.C., 2004
58 CHAPTER 3 DEVELOMENT OF TWO LANE HIGHWAY SIMULAT ION TOOL This chapter describes the development approach for the new two lane highway simulation tool, including the logic and models employed in CORSIM to facilitate two lane highway modeling All of the following tasks were accomplished. Develop traffic flow modeling routines for two lane highway operations. Develop a mechanism for integrating signalized intersections into a two lane highway facility Modify the input and output mechanisms of CORSIM as necessary for the two lane highway modeling capability 3.1 Develop Traffic Flow Modeling Routines for Two Lane Highway Operations The logic of the various components involved i n modeling the two lane highway passing maneuvers (either oncoming lane or passing lane) are described in the following subsections. 3.1.1 Car Following The car following model currently employed in CORSIM i s the Pitt car following model (Halati et al. 19 97) This car following model was also retained for use in the two lane highway modeling. This model incorporates the distance headway and speed differential between the lead and follower vehicle as two independent variables, and the basic assumption is th at the follower vehicle will try to maintain a safe space headway. 3.1.2 Passing in an Oncoming Lane The most distinguishing feature of traffic operations on two lane highways is passing in the oncoming lane (when passing lanes are not present). Therefore, this passing maneuver is constrained by not only the amount of opposing lane distance used in the execution of a passing maneuver, but also the sight distance and clear
59 distance (or gap size) a follower requires before attempting a passing maneuver. The f ormer issue depends on road design and markings of no passing zone, while the latter issue depends on traffic demands. The following subsections describe the various components of logic employed to determine when and how a vehicle will perform a passing ma neuver in the oncoming lane. 22.214.171.124 When will a vehicle attempt to pass a vehicle in front of it? Step 1. Determine if the subject vehicle is in a following mode Currently, the program defines a vehicle as being in a following mode when the time headway between it and the vehicle immediately in front of it is equal to or less than 3 seconds this is currently the logic the HCM 2010 uses to approximate PTSF 1 The value of 3 seconds, however, can be changed by the analyst in the input file. Additionally, t he trailing vehicle must be traveling at a speed at least equal to the speed of the leading vehicle. If the subject vehicle is determined to be in a following mode, then the following steps are carried out to determine if the following vehicle will attempt a passing maneuver. Step 2. Determine tolerable speed If it is determined that a vehicle is in a following mode, then the tolerable speed for that vehicle is calculated. Tolerable speed is defined as the maximum speed at which the desire to pass for a fo llowing driver will be 100 % Tolerable speed varies for different driver types (Eq uation 3 1 for the computation), due to the different degree of aggressiveness for each driver type. Note flow speed and d river type. For 1 It is anticipated that alternative approaches to the determination of whether a vehicle is in a following mode will eventually be considered for incorporation into the program, such as the probabilistic approach by Hideki, et al. ( 200 8).
60 example, a driver of type 1 will have a desired speed of 88% of the link free flow speed, while a driver of type 10 will have a desired speed of 112% of the link free flow speed (these percentages of free flow speed can be modified by users on RT (Record Type) 147 of the TRF input file in CORSIM). ( 3 1 ) where TolerableSpeed i = tolerable speed for driver type i (mi/h) DesiredSpeed i = desired speed for driver type i (mi/h) Step 3. Determine desire to pass The main pass is the difference between their actual travel speed and their desired speed. The 1 based on a non linear function of the cu rrent speed of the subject vehicle. The initial desire to pass ( DTP ) value is given by ( 3 2 ) The DTP is illustrated graphically in Figure 3 1 for one example driver. The DTP value is firstly adjusted by an impatience factor. For vehicles that have a positive DTP but are in a following mode and have not yet initiated a passing maneuver, the impatience factor will incrementally increase the DTP value with each simulation time step; thus increasing the probability of a pa ssing maneuver being initiated. The impatience factor is also a function of the driver type, with larger increments being
61 associated with more aggr essive drivers. The TSWTP counter will initiate when a vehicle has a DTP value greater than zero and will re set when the vehicle completes a pass or its DTP value goes to zero. The impatience value can be adjusted on record type 155. The impatience fact or value is given by DT Value Impatience TSWTP Factor Impatience (3-3) where ImpatienceFactor = the increment to the desire to pass value for each second of time spent wanting to pass TSWTP = time spent wanting to pass (s) DT = driver type (defined by CORSIM) ImpatienceValue = degree of impatience in waiting to pass (default = 0.001) Then the DTP is adjusted based on the length of the trailing vehicle and the length of the leading vehicle. A longer leading vehicle, such as a large truck, will increase the DTP value for the trailing v ehicle. Conversely, the DTP value will be decreased for a longer trailing vehicle. The final adjustment to the DTP value ( AdjDTP ) is given by AdjDTP ( DTP ImpatienceFactor ) 1 1 14 1 FolLeng lnexp(1) 1 14 1 LeadLeng LeadLeng FolLeng 0.5 (3-4) where FolLeng = length of the tr ailing vehicle (ft) LeadLeng = length of the leading vehicle (ft) It should be noted that the value of 14 used in Equation 3-4 is the shortest length of vehicle defined in CORSIM.
62 Finally, if the adjusted DTP value is not less than 0.25, it will be compared to a generated uniform random number between 0 and 1. If the adjusted DTP value is less than the random number, the subject vehicle will continue to follow the leading vehicle. If the adjusted DTP value is greater than or equal to the random number, the subject vehicle will initiate a passing maneuver, subject to other constraints as described in the following section. 126.96.36.199 Constraints governing whether a pass will be initiated If it has been determined that a vehicle wants to initiate a passing maneuver, the following issues are considered. Step 1. Is the vehicle in a passing allowed section? The program logic currently dictates that all passing maneuvers must be initiated in a passing allowed section of the roadway (i.e., skip striping in the applicable direction) 2 However, it is possible for a passing maneuver to be completed in a no passing allowed section, consistent with the field observations from Harwood et al. ( 2008 ) study. This is described in more detail under step 8. Step 2. Check whether a vehicle upstream of the subject vehicle is performing a passing maneuver If the subject vehic le is currently in the process of being passed by another vehicle, the subject vehicle will not initiate its passing maneuver. Step 3. Check whether the maximum number of allowed passing maneuvers is currently in progress The number of vehicles that can b e simultaneously executing a passing maneuver in the oncoming lane is limited to three per each platoon of 2 For computational efficiency reasons, this check is actually performed before the desire to pass calculations.
63 vehicles. Thus, the maximum number of vehicles that can be executing a passing maneuver along the defined length of highway is three times the number of platoons within that defined length of highway. A platoon is defined by a leading vehicle that is not in a following mode and trailing vehicles that are all considered to be in a following mode. Step 4. Check the number of vehicles that must be passed to complete the passing maneuver A vehicle is prevented from starting a passing maneuver when, due to insufficient gaps for merging between the vehicles ahead, there are more than five vehicles that would need to be passed. Step 5. Check whether any queued vehicle exists in the 1 mi downstream segment A vehicle whose speed is equal to or less than 30 ft/s in the 1 mi downstream segment from the subject vehicle is considered to be queued. If any queued vehicle is found, the subject vehicle will not initiate its passing maneuver. Step 6. Determine the required passing sight distance If the subject vehicle is allowed to initiate a passing maneuver per the above constraints, then the passing sight distance ( PSD ) will be calculated. PSD is the minimum distance necessary between the potential passing vehicle and an oncoming vehicle that will still allow the potential passing vehicle to safely initiate and complete a passing maneuver of a leading slower vehicle in the oncoming lane. If the horizontal and/or vertical alignment aspects of a highway do not provide unobstructed sight distance at least equal in length to the PSD then the highway is typically striped with solid yellow center lines (i.e., no passing
64 allowed) 3 Both A policy on Ge ometric Design of Highways and Streets (AASHTO 2004) and the Manual on Uniform Traffic Control Devices fo r Streets and Highways (MUTCD) (FHWA 2003) provide minimum PSD values for design and marking. However, their respective recommended PSD values vary s ignificantly, with the AASHTO recommended values being considerably larger than the MUTCD recommended values. Harwood et al. (2008) proposed recommendations on the adequacy of current procedures and guidelines used to estimate minimum PSD requirements for highway design and pavement marking. Based on the literature review and t he field study, Harwood et al. (2008) in dicated that the Glennon model (1988) and the Hassan et al. model (1996) are the most reasonable models, which result in considerably shorter PSD values than the AASHTO recommended values, much closer to the MUTCD recommended values. Furthermore, the Harwood et al. study argued that crash statistics for two lane highways do not provide much support for the notion that the current practice for ma rking passing zones (based on the MUTCD criteria) is in need of revising. Based on the field data results of the Harwood et al. (2008) study, the authors recommend the following assumptions be used when applying the Glennon and Hassan et al. models: The sp eeds of the passing and opposing vehicles are equal and represent the design speed of the highway. 3 CORSIM does not compute available passing sight distance based on the specified roadway geometry; thus, it is necessary for the user to explicitly identify the a llowable passing zones along the length of the highway.
65 The vehicle being passed travels at uniform speed and the speed difference between the passing vehicle and vehicle being passed is 12 mi/h. The passing vehi cle has sufficient acceleration capability to reach the specified speed difference relative to the passed vehicle by the time it reaches the critical position (same as the assumption that each vehicle will travel at the constant speed after the critical po sition) The lengths of the passing and passed vehicles are 19 ft. reaction time in deciding to abort a passing vehicle is 1 sec ond The deceleration rate used in aborting a pass is 11.1 ft/s 2 For a completed or aborted pas s, the headway between the passing and passed vehicles is 1 sec ond The minimum clearance headway between the passing and opposing vehicles at the point at which the passing vehicle returns to its normal lane is 1 sec ond By using these assumptions, the Gl ennon and Hassan et al. models result in PSD values similar to the MUTCD recommended PSD values. The default PSD model used in CORSIM is the AASHTO model ( Table 3 1 and Figure 3 2 for the details). However, it is also possible to use the MUTCD PSD values i nstead. Users can modify several CORSIM inputs to implement the MUTCD PSD values ( Table 3 2 for the revised values of the variables). If the available sight distance for the subject vehicle is equal to or greater than the minimum required PSD value, the fo llowing check is made. Step 7. Check length of passing zone and compare to the minimum passing zone length The length of the passing zone (as indicated by roadway markings) is compared to the minimum passing zone length (which is equal to distance travele d by the passer before it returns to the normal lane during the passing maneuver, as suggested by AASHTO). A vehicle will not initiate a passing maneuver unless the
66 available passing zone length is equal to or greater than the minimum passing zone length. If the marked passing zone length is greater than the minimum passing zone length, then the following check is made. Step 8. Determine effective passing zone length and compare to the distance needed to complete the pass As mentioned earlier, it is possib le for a vehicle to complete its passing maneuver in a section of roadway marked as no passing allowed. The length of the available passing zone (from the current position of the passing vehicle) is initially determined from just the roadway markings. This value is then adjusted based on the permissible amount of distance beyond the marked passing zone length, calculated as follows: (3 5 ) where IllegalPassDistPct = allowable percentage of the total passing distance beyond the passing zone (as indicated by roadway markings) MinPct = allowable percentage of the total passing distance corresponding to vehicle of driver type 1 (default = 0) MaxPct = allowable percentage of the total passing distance corresponding to vehicle of driver type 10 (default =25)
67 ( 3 6 ) where DistAvailForPass = the total amount of distance available for passing, based on roadway markings and allowable distance downstream of marked passing zone AvailPassZoneDist = length of marked passing zone available a t beginning of passing maneuver The distance needed to complete the pass ( DNTCP ), which before the passing maneuver is initiated is equal to the estimated distance traveled by the subject vehicle during its passing maneuver (in the PSD calculations in step 6), is compared to the distance available for the pass (i.e., effective passing zone length). If the DNTCP is less than the effective passin g zone length, then the passing maneuver can be initiated. 188.8.131.52 How a potential passer executes its passing maneuver? If all the requirements discussed in the previous section are satisfied for a potential passing vehicle, it will initiate the passing m aneuver. The general logic of executing a passing maneuver is divided into three stages, as illustrated in Figure 3 3. Stage 1: Initiate passing in the normal lane At the beginning of a passing maneuver, the potential passer starts to accelerate at the ac celeration rate based on its own speed according to the AASHTO criteria or revised AASHTO criteria for approximating the MUTCD criteria. At the same time, the potential passer moves over to the opposing lane. Stage2: Passing in the opposing lane After the potential passer moves into the opposing lane, it will keep on accelerating until it reaches a speed 12 mi/h (this value
68 can be changed on RT 155) greater than the speed of the vehicle being passed 4 Meanwhile, the variable DNTCP (distance needed to complete pass) is compared to the variable DTC 5 (distance to collision with the oncoming vehicle) every time step (Figure 3 4 for an illustration of these variables). For the subject vehicle who finds that more than one vehicle need to pass before it initiates the passing maneuver, the DNTCP is based on the position of the most downstream vehicle to be passed. As mentioned previously, before the passing maneuver is initiated, the DNTCP is equal to the estimated distance traveled by the subject vehicle during its passing maneuver (in the PSD calculations in step 6). However, once the passing maneuver is in progress, the DNTCP is continually changing (generally decreasing). For calculating DNTCP the time needed to complete passing ( TNTCP ) is first calculated. Since the potential passer will stop accelerating when its speed is 12 mi/h (i.e., 17.6 ft/s) greater than the vehicle being passed, the value of TNTCP is computed ( 3 7 ) ( 3 8 ) ( 3 9 ) where TNTCP = time n eeded to complete passing (s) t accel = time needed to reac h the maximum passing speed (s) 4 Based on NCHRP R eport 605 (Harwood et al, 2008). 5 The DTC will be infinite when no oncoming vehicle is present.
69 t const = time needed to complete passing after the accelerating period (s) t 0 = time step when the passing maneuver initiates (s) t = current time step (s) v 1 (0) = initial spee d of the passing vehicle (ft/s) v 2 (0) = initial speed of the vehicle being passed (ft/s) m = speed differential between the passing vehicle and the vehicle being passed when the passing vehicle re aches the maximum passing speed a 1 = acceleration of the passing vehicle (ft/s 2 ) VehPos 1 = position of the passing vehicle relative to the upstream en d of the segment when it stops accelerating (ft) VehPos 2 = position of the vehicle being passed relative to the upstream end of the segment when the passing vehicle stops accelerating (ft) Once the TNTCP is obtained, the DNTCP can be calculated by ( 3 10 ) where v 1 ( t ) = speed of the passing vehicle at time step t (ft/s) v 2 ( t ) = speed of the vehicle being passed at time step t (ft/s)
70 ( 3 11 ) where VehPos 3 = position of the oncoming vehicle relative to the upstream end of the segment at time step t (ft) v 3 (t) = speed of the oncoming vehicle at time step t (ft/s) After the subject vehicle moves onto the opposing lane to perform its passing maneuver, the value of DNTCP is compared to the value of DTC at every time step until the passing maneuver is completed. The result of this comparison leads to the following different situations that must be considered. If the DNTCP is less than the DTC the passer will conti nue its passing maneuver as planned. Another issue related to the completion of passing is the gap size in front of the vehicle being passed, which is included in the calculation of DNTCP The passer requires a certain gap size to be able to return to the normal lane in order to complete passing. The default value for this gap size (i.e., ClearGap in Figure 3 4) is 75 ft, but can be revised on Record Type 155. Normally, the passing vehicle only accepts the gap in front of the vehicle being passed if it is g reater than or equal to the minimum gap. In some cases, the passing vehicle will consider to pass the next vehicle if the gap in front of the vehicle being passed is not sufficient. However, if the passing vehicle has already passed 5 vehicles, it will be forced to squeeze back to the normal lane. I f the DNTCP is greater than or equal to the DTC the decision on whether the passing vehicle will continue or abort its passing maneuver is dependent on the relative position of the passing vehicle to the passed vehicle. A study by Harwood and Glennon ( 1976 ) defined the critical position as the point a t which the sight distances required to
71 abort the pass and to complete the pass are equal. Two studies, one by Vanvalkenburg et al. ( 1971 ) and one by Weaver et al. ( 1972 ) independently recognized a key position of a passing maneuver occurs at the point whe re the passing driver can no longer safely abort the pass and is, therefore, committed to complete it. However, no unanimously accepted definition of the critical position currently exists. Harwood et al. ( 2008 ) suggests that the critical position can be c onsidered to be the point when the passing vehicle and vehicle being passed are directly abreast of one another. This definition is applied in the CORSIM logic. Two situations were identified as follows: The passer has not reached the critical position If the passing vehicle has not yet reached the critical position when DNTCP is greater than DTC it will abort the passing maneuver. The process of returning to the normal lane is similar to that of completing a passing maneuver. The gap behind the vehicle b eing passed will be checked for the return. The minimum gap here is also set as three times the length of the passing vehicle, and the passing vehicle will only accept a gap greater than or equal to the minimum gap. The passer has reached the critical posi tion If the passer has reached the critical position when DNTCP is greater than DTC the passer will continue with completing the passing maneuver. In order to accommodate this without collision, the passing vehicle will speed up and/or the oncoming vehic le will decelerate. The specific amount of acceleration by the passing vehicle and/or deceleration by the oncoming vehicle is a function of the current acceleration rate of each vehicle, the capabilities, and the current DTC value. In the simulation, the vehicle being passed is set to be aware of this expedited passing, and it usually slows down to assist the passing maneuver. Occasionally, the vehicle being passed may not slow down even though it is aware of the situation, in such cases, the passing vehicle will be forced to abort its passing maneuver. Stage3. Return to the normal lane For a passing maneuver being completed, the passer will return to the normal lane in front o f the vehicle being passed when the gap is sufficient. For an aborted passing maneuver, the passer will return to the normal lane behind the vehicle being passed if there is a sufficient gap. The existing mandatory lane changing logic in CORSIM is utilized for this situation.
72 Under certain conditions, the passing vehicle will consider passing more than one vehicle in front of the current vehicle being passed, and the gap in front of the current vehicle being passed is insufficient, the passing vehicle will attempt to pass the vehicle in front of the current vehicle being passed, subject to the logic and constraints as previously discussed. 3.1.3 Passing in a Passing l ane S ection A passing lane is defined as a lane added to improve passing opportunities in one direction of travel on a conventional two lane highway. Although it may vary by jurisdiction, the logic implemented in this simulation program assumes that slower vehi cles will move to the right lane in a passing lane section and the passing vehicles reco gnized that this does not always happen; thus, the developed logic allows for the possibility that an impeding vehicle will not move over. For each vehicle in a passing lane section (hereafter referred to as the subject vehicle), the logic first checks th e headway between this subject vehicle and the vehicle immediately behind, and then checks the headway between this subject vehicle and the vehicle immediately ahead. If the subject vehicle has a vehicle behind it in following r headway threshold) and is not in following mode itself, the willingness to move over ( WTMO ) to the right lane of the passing lane section for the subject vehicle will be considered. The value of WTMO is determined as follows:
73 ( 3 12 ) where WTM O = willingness to move over SVS = s ubject vehicle speed (mi/h) FFS = free flow speed (mi/h) This parameter will be adjusted by dividing by the square root of driver type if the length of the subject vehicle is less than 40 ft (the length of a single unit truck), otherwise it remains as the original value (not dividing by the square root of driver type results in a higher probability to move over for trucks). The adjusted willingness to move over is compared to a generated uniform random numbe r between 0 and 1, and if it is greater than the random number the subject vehicle will move over to the right lane; otherwise, it will stay in the left lane. The existing CORSIM logic is utilized for determining when a vehicle will move from the right lan e (which drops at the end of the passing lane section ) back to the normal lane. Generally, the discretionary lane change logic will apply for vehicles in the right lane (either slow vehicles that moved out of the way, or faster vehicles that are trying pas s slower vehicles that did not move over) until such a vehicle gets near to the end of the right lane, in which case the mandatory lane change logic will be applied. 3.2 Integrating Signalized Intersections into a Two Lane Highway Facility CORSIM consists of two microscopic stochastic simulation models: 1) NETSIM, which models traffic on urban streets, including intersections, and 2) FRESIM, which
74 models traffic on freeways. The previous two lane highway modeling features were built into the FRESIM model. Since the capability to connect FRESIM links to NETSIM links already existed, it was not necessary to make any special modifications to the underlying CORSIM models to accommodate the modeling of signalized intersections within a two lane hi ghway. Thus, the rest of this section simply provides guidelines for defining two lane highway networks that contain signalized intersections, particularly the interface (i.e., connection) between the two lane highway links and the signalized intersection links. The signalized intersection and two lane highway features should be defined as usual in the TRF file. Then, according to the CORSIM node numbering scheme 6 two nodes numbered between 7000 and 7999 should be added along the two lane highway direction as the exit nodes for the NETSIM links, and also the entry nodes for the no less than 100 feet (Figure 3 5), since some vehicles with high speeds could possibly passing zone markings extend at least 200 feet upstream and downstream of the intersection (DelDOT 2009). Passing in the opposing lane is not allowed on NETSIM links; thus, if the NETSIM link s are less than 200 feet in length, no passing zone markings should be specified for the adjoining FRESIM links. 6 Node numbers 1 through 6999 can be used for internal nodes. Node numbers 7000 through 7999 can be used for interface nodes and node numbers 8000 through 8999 can be used for entry or exit nodes.
75 3. 3 Inputs and Outputs of the Simulation Program 3.3.1 New I nputs Several new inputs were added to CORSIM to accommodate user revision of vari ous parameters for the two lane highway modeling logic. The new inputs generally fall into two categories as follows: F acility property inputs This type of inputs allows users to create a two lane highway segment and make the appropriate marking configurations on it. Users can also define a two lane highway by direction, by specifying the links that comprise the highway direction. Passing maneuver inputs This type of inp uts a llows users to modify the values of the components in the passing models, such as follower headway threshold, speed differential between passing vehicle and passed vehicle or the inputs used to calculate the minimum passing sight distance required. T he details of the new inputs are listed in Table 3 3. 3.3.2 New O utputs Several new outputs were added to CORSIM to accommodate a variety of new performance measures and passing lane diagnostics. 184.108.40.206 Performance measure outputs In the HCM 2010, the level of service on a two lane highway is defined based on both PTSF and ATS In the simulation, these two measures can be selected as outputs follower density is added to the CORS IM outputs. According to the definition, follow density can be calculated by ( 3 13 ) All the three performance measures can be output in the simulation program either on link basis or on facility basis. Facility wide values of ATS PT SF and follower density are included in the default output file (.out) of CORSIM.
76 In addition, another potential performance measure percent delay ( PD ) is also added as one of the outputs, for developing the analytical methodology in Chapter 4. The value of PD is determined by ( 3 14 ) where DelayTravelTotal = t otal time that all vehicles were delayed in the network (CORSIM definition) MoveTimeTotal = t otal theoretical time for discharged vehicles to travel the length of the link if moving unimpeded at the free flow speed (CORSIM definition) 220.127.116.11 Passing statistics outputs As passing is the most significant feature of two lane highway operations, the outputs of passing statistics can be employed as the supplementary information for analyzing two lane highway operations. Users can get the number of attempted passes, the number of aborted numbers, and the number of completed passes in a specific simulation run. Other information such as the distance traveled by the passer in the oppos ing lane during the simulation time period can be obtained as well. Most of these outputs are specified through the output processor tool in TSIS. The can then be loaded into a spreadsheet program. The details of the new outputs are listed in Table 3 4.
77 3. 4 Preliminary Tests on t he New Simulation Program The new simulation program is tested by running a set of scenarios, to verify whether the simulation results are consistent with commonly accepted traffic flow theories. A large number of experiments were conducted for the purpose of comparing the simulation results to the traffic flow theories for two lane highways. In these experiments, follower density, PTSF and ATS were chosen as the primary performance measure outputs. 3. 4 .1 Experimental Design I t is expected that not only the traffic operations on basic two lane highways are simulated and tested, but also the traffic operations on two lane highways with various features such as passing lanes and signalized intersections. Besides the different facility geometries for different sets of experiments (Figure 3 6 for configuration of each facility) the appropriate options of two way flow volume, directional volume sp lit and percentage of heavy vehicles are applied to each experiment as listed in Tables 3 5 and 3 6 For all the experiments, 10 iterations of each experiment were run, and the average simulation results of the 10 iterations were obtained and analyzed. The findings from the data analysis will be discussed in the following sections, as well as the conclusions for the preliminary tests. 3. 4 .2 Findings of Data Analysis For capturing the difference of performance measures resulted by different features on tw o lane highways, the findings of the data will be summarized for the basic segment tests, passing lane tests and signalized intersection tests, respectively.
78 3. 4 .2.1 Basic segment tests The tests in this part uses basic two lane highway segment as the tes ting facility, without any special features. The only difference regarding the facility geometry is the center line markings, which determine the lengths of passing zones. Therefore, the performance measure results for different passing zone conditions wil l be compared with each other, and presented in Figure 3 7 to 3 9. Because only 0% heavy vehicles was used in the basic segment tests with 50% passing zone, for the convenience of comparing the results between different sets of basic segment tests, only th e scenarios without trucks are included in this part. Directional average speed vs. volume Figure 3 7 illustrates the relationship between directional average speed and two way flow rate. These results follow the expected trends and relationships: W hen th e t raffic flow rates are approximately equal on each lane of the two lane highway, the average speed for each direction is similar; as the traffic directional split becomes more imbalanced, the average speeds in the peak direction are lower than in the off peak direction, and the average speeds decline with increasing flow rates in each direction T he average speeds are higher for the 100% passing allowed condition than the no passing allowed condition, for each flow rate scenario, until you get to high flow rates (approximately 2 500 veh/h two way flow rate), at which point the average speeds are similar High flow rates severely limit the passing opportunities; thus, essentially creating a no passing allowed condition Also, the improvement to peak dire ction average speed for the passing allowed condition relative to the no passing allowed condition increases with increasing traffic directional split imbalance, as expected Also
79 as expected, whether for the passing allowed or no passing allowed condition the average speed levels off in the moderate to high two way flow rate range As to the 50 % passing allowed condition the performance f or the EB direction is initially (i.e., at lower flow rates) in between the performance of the two 100 % conditions un til it converges to the 100% no passing allowed values, which it does at a lower flow rate than the 100 % passing allowed condition. This suggests that the effect of proportion of passing zone on the performance measures is not linear. For the WB direction, the average speeds for the 50 % passing allowed condition generally follow those for the 100 % passing allowed condition. This may seem to be a somewhat unexpected result, but it is plausible. The passing logic allows a vehicle to complete its pass a certai n distance downstream of the end of the passing zone (as mentioned in section 3. 1 .2.2). The passing logic also requires a passing maneuver to be completed before it exits the coded highway facility. For the facility configuration shown in Figure 3 6C pass ing vehicles traveling in the EB direction must complete their pass before reaching the end of the 5 mi passing zone; whereas passing vehicles traveling in the WB direction can complete their pass partially into the no passing allowed zone. Thus, the effec tive passing zone length is longer for the WB direction than it is for the EB direction, which explains why the 50 % passing allowed results are better in the WB direction than in the EB direction. Directional PTSF vs. volume Figure 3 8 illustrates the rel ationship between directional PTSF and two way flow rate. The results of the relationship between the PTSF values and flow rate are also as expected. When the traffic flow rates are approximately equal on each lane of the two lane highway, the PTSF value f or each
80 direction is similar; as the traffic directional split becomes uneven, the PTSF in the peak direction is higher than in the off peak direction, and the PTSF value increases with increasing flow rates in each direction. The PTSF values are lower for the 100% passing allowed condition than the no passing allowed condition, for each flow rate scenario, until the two way flow rates reaches approximately 2,500 veh/h, where the PTSF values are similar. This is consistent with the resu lts for average speed. And the improvement to peak direction PTSF for the passing allowed condition relative to the no passing allowed condition increases with increasing traffic directional split imbalance. Also as expected, whether for the passing allowe d or no passing allowed condition, the PTSF levels off in the moderate to high two way flow rate range. Similar to the average speed results, for the EB direction, the PTSF results of the 50 % passing allowed condition is initially in between the performance of the other two conditions until it converges to the 100 % no passing allowed values. And for the WB direction, the PTSF values for the 50 % passing allowed condition generally follow those for the 100 % passing allowed condition. Directional follower density vs. volume Figure 3 9 illustrates the relationship between directional follower density and two way flow rate. The results of the relationship between the follower density values and fl ow rate are as expected: When the traffic flow rates are equally split on each lane of the two lane highway, the follower density for each direction is similar; as the traffic directional split becomes more imbalanced, the follower density in the peak dire ction is higher than in the off peak direction, and increases with increasing flow rates in each direction.
81 The follower densities are lower for the 100% passing allowed condition than the no passing allowed condition. For each flow rate scenario, until yo u get to high flow rates (approximately 2,500 veh/h two way flow rate), at which point the follower densities become similar. Again, this is because high flow rates severely limit the passing opportunities, essentially creating a no passing allowed conditi on. Also, the improvement to peak direction follower density for the passing allowe d condition relative to the no passing allowed condition increases with increasing traffic directional split imbalance, as expected. As to the comparison between the 50% pas sing allowed conditions and the other two conditions, t he results for the follower density values follow the same trends and relationships as described for the average speed results. 3. 4 .2. 2 Two lane highway with a passing lane tests As passing lanes are d esigned for improving passing opportunities by providing slow vehicles with an extra lane to move over, more slow vehicles are expected in passing lane tests to better reflect the benefits of adding a passing lane. Therefore, 10% of traffic flow is set as trucks, and the results are compared with the ones for the basic segment tests with no passing zones Directional average speed vs. volume Figure 3 10 illustrates the relationship between directional average speed and two way flow rate. The results for average speed generally follow the expected trends and relationships. In the EB direction (where the passing lane is present), the average speeds are generally higher than the speeds for the 100 % no passing allowed and no passing lane condition. The averag e speeds for the WB direction (with no passing lane) are similar to the speeds for the 100 % no passing allowed condition. The results
82 indicate that the improvement to average speed will diminish around 2,500 veh/h. This is because at high flow rates, there is enough turbulence (particularly from trucks) created at the merge point at the end of the passing lane to offset the addition of the passing lane. The WB results show that the passing lane added in the EB direction does not offer any speed benefit to t he WB direction, which is as expected. Other trends and relationships in the average speed results are as previously described for the other average speed results. Directional PTSF vs. volume Figure 3 11 illustrates the relationship between directional PT SF and two way flow rate. The presence of the passing lane generally provides similar improvements to the PTSF values in the EB direction as found for the average speed results. Likewise, for the WB direction (with no passing lane), the PTSF results are vi rtually identical. Directional follower density vs. volume Figure 3 12 illustrates the relationship between directional follower density and two way flow rate. The presence of the passing lane generally provides similar improvements to the follower densit y values in the EB direction as found for the average speed and PTSF results. Likewise, for the WB direction (with no passing lane), the follower density results are virtually identical. 3. 4 .2. 3 Two lane highway segment with a signalized intersection test A test on a two lane highway segment with a signalized inters ection is also done to see the impact of adding a signalized interse ction. As shown in Figure 3 6E p assing is not allowed in the NETSIM links used for modeling the signalized intersection operations. Additionally, passing is not allowed on the adjacent upstream and downstream 600 foot segments. Passing is allowed on a ll other segments. The flow rate
83 for the major highway is 5 00 veh/h (no trucks) for each direction, and the flow rate for the minor crossing street is 100 veh/h (no trucks) for each direction. The cycle length is 100 seconds Only one experiment w as run for this configuration, just to demonstrate that the effect of the signalized intersection is reflected in the facility perform ance measure results. The testing results produced by CORSIM were compared with the ones with the basic two lane highway ( Table 3 7 ). The results confirm the expectation that the presence of a signalized intersection will worsen the operational performance on the whole facility. Additional investigation of the effects of signalized intersections along a two lane highway will be presented in Chapter 4 3. 4 3 Conclusions of the Preliminary Tests The results of the preliminary tests indicate that the new simul ation capability in CORSIM can model traffic operations on complex two lane highways (e.g., two lane highway with occasional signalized intersections). I t is also indicated that the two lane modeling logic applied in the simulation program is reasonably co nsistent with theories and field observations discussed in the literature, and the modeling results are generally reasonable and consistent with expected traffic flow theory for two lane highways
84 Table 3 1. Components and values of passing sight distance (AASHTO 2004 ) Component of passing maneuver Speed range (mi/h) 30 40 40 50 50 60 60 70 Average passing speed (mi/h) 34.9 43.8 52.6 62.0 Initial maneuver: + a 1.40 1.43 1.47 1.50 ++ t 1 3.6 4.0 4.3 4.5 +++ d 1 Occupation of left lane: ** t 2 9.9 7 *** d 2 1.467 # v t 2 Clearance distance: ## d 3 100 180 250 300 Opposing vehicle: ### d 4 0.667 d 2 Minimum PSD d 1 + d 2 + d 3 + d 4 + acceleration of passing vehicle when initiating passing maneuver (mi/h/s) ++ time required for initial maneuver (s) +++ initial maneuver distance traveled from start of passing maneuver until passing vehicle encroaches upon oncoming lane (ft) *difference in speed of passed vehicle and passing vehicle ( m i/h) **t ime spent traveling in the oncoming lane (s) ***the distance traveled by the passing vehicle from the point of encroachment in the oncoming lane to the point of return to the normal lane (ft) # average speed of passing vehicle ( m i/h) ## t he shortest desirabl e distance between the front bumpers of the passing and opposing vehicle when the passing vehicle returns to the normal lane (ft) ### the distance traveled by the opposing vehicle during the time the passing vehicle travels from the position of being direct ly abreast of the vehicle being passed to the return to the normal lane (ft) 7 The original AASHTO criteria included four separate values for t 2 based on the passing vehicle speed. The Harw ood et al. (2008) study recommended a constant value of t 2 because no evidence has been found that the left lane travel time of a passing vehicle increases as the speed of the passed vehicle increases The chosen value was the mean value of 9.9 seconds from their study.
85 Table 3 2 Revised values of the AASHTO PSD model variables for approximating the MUTCD PSD values Component of passing maneuver Speed range (mi/h) 30 40 40 50 50 60 60 70 Average passing speed (mi/h) 34.9 43.8 52.6 62.0 Initial maneuver: t 1 3. 0 2.5 2.0 1. 6 Occupation of left lane: t 2 5.9 6.0 6.2 6.4 Clearance distance : d 3 80 100 120 140
86 Table 3 3 New inputs for two lane highway simulation Parameter Description Range Default Field Record Type Passing Zone (See Appendix A for the corresponding marking configuration) Specifies the center line marking of a two lane highway segment. 0 Not Applicable 1 Passing Not Allowed 2 Passing Allowed 3 Passing Allowed in Both Directions 0 74 20 Entry node of eastbound segment of a two lane highway Specifies the two node number (for EB). 7000 8999 none 1 4 154 Entry node of westbound segment of a two lane highway Specifies the two node number (for WB). 7000 8999 none 5 8 154 Follower Headway Threshold ( s ) Specifies the follower headway threshold in seconds. It is used to set the time headway value that is used to determine whether a vehi cle is in a following mode. 1.0 10.0 3 1 4 155 MinPct (%) Specifies the illegal passing distance percentage acceptable for driver type 1 (most conservative driver). 0 100 0 5 8 155 MaxPct (%) Specifies the illegal passing distance percentage acceptable for driver type 10 (most aggressive driver). 0 100 25 9 12 155 Minimum clearance distance between passing vehicle and passed vehicle (ft) Specifies the shortest distance between the front bumper of the passed vehicle and the rear bumper of the passing vehicle, when the passing vehicle moves back to the normal lane, which allows the passing vehicle to safely complete its pass. 10 100 75 13 16 155 TimeToCollisionMin ( s ) Specifies the time to collision th at will cause a type 10 driver to abandon passing an additional vehicle during passing maneuver. 0 100 5 17 20 155
87 Table 3 3. Continued Parameter Description Range Default Field Record Type TimeToCollisionMax ( s ) Specifies the time to collision that will cause a type 1 driver to abandon passing an additional vehicle during passing maneuver. 0 100 10 21 24 155 Speed differential between passing vehicle and passed vehicle (mi/h) Specifies the difference in speed between the passing vehicles and vehicle being passed when the passing vehicle has reached its top passing speed. 5 20 12 25 28 155 Impatience Value Specifies the increment to the desire to pass value for each second of time spent wanting to pass. 0.000 0.001 0.001 61 64 155 Passing acceleration Specifies the acceleration that will be applied to accelerate the passing vehicle to the desired passing speed when a passing maneuver is initiated. a 1 a 2 a 3 a 4 if speed > 60 mi/h a 1 = 1.40 mi/h/s a 2 = 1.43 mi/h/s a 3 = 1.47 mi/h/s a 4 = 1.50 mi/h/s 1 16 156 Time required for initial passing maneuver Specifies the amount of time the passing vehicle spends moving from the normal lane to the point of encroachment in the oncoming lane t 1 t 2 t 3 t 4 if speed > 60 mi/h t 1 = 3.6 s t 2 = 4.0 s t 3 = 4.3 s t 4 = 4.5 s 17 32 156 Time spent traveling in the oncoming lane Specifies the amount of time the passing vehicle spends from the point of encroachment in the oncoming lane until it returns to the normal lane. t 1 t 2 t 3 t 4 if speed > 60 mi/h t 1 = 9.9 s t 2 = 9.9 s t 3 = 9.9 s t 4 = 9.9 s 33 48 156 Minimum clearance distance Specifies the minimum acceptable clearance distance between the passing vehicle and the first oncoming vehicle. c 1 c 2 c 3 c 4 if speed > 60 mi/h c 1 = 100 f t c 2 = 180 ft c 3 = 250 ft c 4 = 300 ft 49 64 156 Two Lane Highways by Direction Allows the user to define a two lane highway by direction N/A N/A 190 Two Lane Highway Direction Links Used in conjunction with RT 190 to specify the links that comprise the highway direction N/A N/A 191
88 Table 3 4 New o utputs for t wo lane h ighway s imulation (by direction) Output Description Formula Facility based or link based Follower density Indicates the average density of vehicles which are in the following mode. Percentage of followers Traffic density (Followers per mile) Both Percent time spent following Indicates the average percentage of travel time spent by vehicles following in platoons behind slower vehicles. (%) Both Average travel speed Indicates the average travel speed of vehicles. (mi/h) Both Delay Travel Per Move Time Ratio ( PD ) Indicates the ratio of delay to free flow travel time. (%) Both Average speed of vehicles while performing a passing maneuver Indicates the average travel speed of all the passers during their processes of passing maneuvers (i.e. from initiating passing to returning to the normal lane) (mi/h) Facility Number of attempted passes Indicates the total amount of the initiated passing maneuvers, including both completed and aborted passes over the whole facility. N/A Facility Number of aborted passes Indicates the total number of passes that are initiated but not completed. N/A Facility Clearance distance and/or time Indicates the average clearance (in terms of distance or time) between passing vehicle and opposing vehicle when the passer completes the passing maneuver and returns to the normal lane. (ft or s ) Facility
89 Table 3 4. Continued Output Description Formula Facility based or link based Average distance spent in opposing lane Indicates the average distance the passer travels in opposing lane during the passing maneuver. (ft) Facility Average proportion of illegal passing distance Indicates the average proportion of passing distance used beyond a passing zone, relative to the total passing distance. (%) Facility
90 Table 3 5 Preliminary e xperimental design (1) facility geometry Facility length (mi) % Passing zone Terrain Other features Basic segment tests 10 100 Level N/A 10 0 Level N/A 10 50 Level N/A Passing lane tests 10 0 Level A EB passing lane added from milepost 4 to 5 Signalized intersection test 3 0 on the adjacent upstream and downstream 600 f t segments 100 on others Level A signalized intersection added in the middle Table 3 6 Preliminary e xperimental design (2) traffic inputs Option 1 2 3 4 5 6 7 8 9 Two way volume (veh/h ) 100 500 1000 1500 2000 2500 3000 3500 4000 Directional splits (EB/WB) 50/50 60/40 ** 70/30 % Heavy vehicle *** 0 **** 10 Free flow speed (mi/h) 65 This split does no t apply for two wa y volume of 3500 and 4000 veh/h **This split does not apply for two way volu me of 3000, 3500 and 4000 veh/h *** This percentage does not apply for the passing lane tests. **** This percentage does not apply for the basic segment tests with 50% passing zone Table 3 7 Two lane highway with a signalized intersection testing results Scenario Avg. Speed (mi/h) PTSF (%) Follower Density (veh/mi) Signalized intersection added EB 54.70 46.38 1.53 WB 54.67 46.87 1.56 No signalized intersection EB 62.27 34.58 0.62 WB 62.20 35.03 0.65
91 Figure 3 1 An e xample of Desire to Pass for a specific driver Figure 3 2 Elements of passing sight distance for two lane highways (AASHTO 2004 )
92 Figure 3 3 General passing maneuver process
93 Figure 3 4 Illustration of DNTCP and DTC Figure 3 5 Example of the interface between FRESIM links and the NETSIM links
94 A B C D Figure 3 6. Configurations of testing facilities. A) B asic segment tests with100% passing zone B) Basic segment tests with no passing zone C) Tests with 50% passing zone D) Tests with a passing lane E) Test with a signalized intersection
95 E Figure 3 6 Continued
96 A B C Figure 3 7 Directional average speed results for basic segment tests A) EB Avg. Speed, 50/50 Volume Split B) WB Avg. Speed, 50/50 Volume Split C) EB Avg. Speed, 60/40 Volume Split D) WB Avg. Speed, 4 0/ 6 0 Volume Split E) EB Avg. Speed, 70/30 Volume Split F) WB Avg. Speed, 3 0/ 70 Volume Split 56 57 58 59 60 61 62 63 64 65 66 0 500 1000 1500 2000 2500 3000 3500 4000 Average Travel Speed (mi/h) Volume (veh/h) 50% no-passing-allowed and 50% passing-allowed 100% no-passing-allowed 100% passing-allowed 56 57 58 59 60 61 62 63 64 65 66 0 500 1000 1500 2000 2500 3000 3500 4000 Average Travel Speed (mi/h) Volume (veh/h) 50% no-passing-allowed and 50% passing-allowed 100% no-passing-allowed 100% passing-allowed 56 57 58 59 60 61 62 63 64 65 66 0 500 1000 1500 2000 2500 3000 3500 4000 Average Travel Speed (mi/h) Volume (veh/h) 50% no-passing-allowed and 50% passing-allowed 100% no-passing-allowed 100% passing-allowed
97 D E F Figure 3 7 Continued 56 57 58 59 60 61 62 63 64 65 66 0 500 1000 1500 2000 2500 3000 3500 4000 Average Travel Speed (mi/h) Volume (veh/h) 50% no-passing-allowed and 50% passing-allowed 100% no-passing-allowed 100% passing-allowed 56 57 58 59 60 61 62 63 64 65 66 0 500 1000 1500 2000 2500 3000 3500 4000 Average Travel Speed (mi/h) Volume (veh/h) 50% no-passing-allowed and 50% passing-allowed 100% no-passing-allowed 100% passing-allowed 56 57 58 59 60 61 62 63 64 65 66 0 500 1000 1500 2000 2500 3000 3500 4000 Average Travel Speed (mi/h) Volume (veh/h) 50% no-passing-allowed and 50% passing-allowed 100% no-passing-allowed 100% passing-allowed
98 A B C Figure 3 8 Directional PTSF results for basic segment tests A) EB PTSF 50/50 Volume Split B) WB PTSF 50/50 Volume Split C) EB PTSF 60/40 Volume Split D) WB PTSF 4 0/ 6 0 Volume Split E) EB PTSF 70/30 Volume Split F) WB PTSF 3 0/ 70 Volume Split 0 10 20 30 40 50 60 70 80 90 100 0 500 1000 1500 2000 2500 3000 3500 4000 Percent Time Spent Following (%) Volume (veh/h) 50% no-passing-allowed and 50% passing-allowed 100% no-passing-allowed 100% passing-allowed 0 10 20 30 40 50 60 70 80 90 100 0 500 1000 1500 2000 2500 3000 3500 4000 Percent Time Spent Following (%) Volume (veh/h) 50% no-passing-allowed and 50% passing-allowed 100% no-passing-allowed 100% passing-allowed 0 10 20 30 40 50 60 70 80 90 100 0 500 1000 1500 2000 2500 3000 3500 4000 Percent Time Spent Following (%) Volume (veh/h) 50% no-passing-allowed and 50% passing-allowed 100% no-passing-allowed 100% passing-allowed
99 D E F Figure 3 8 Continued 0 10 20 30 40 50 60 70 80 90 100 0 500 1000 1500 2000 2500 3000 3500 4000 Percent Time Spent Following (%) Volume (veh/h) 50% no-passing-allowed and 50% passing-allowed 100% no-passing-allowed 100% passing-allowed 0 10 20 30 40 50 60 70 80 90 100 0 500 1000 1500 2000 2500 3000 3500 4000 Percent Time Spent Following (%) Volume (veh/h) 50% no-passing-allowed and 50% passing-allowed 100% no-passing-allowed 100% passing-allowed 0 10 20 30 40 50 60 70 80 90 100 0 500 1000 1500 2000 2500 3000 3500 4000 Percent Time Spent Following (%) Volume (veh/h) 50% no-passing-allowed and 50% passing-allowed 100% no-passing-allowed 100% passing-allowed
100 A B C Figure 3 9 Directional follower density results for basic segment tests A) EB follower density, 50/50 Volume Split B) WB follower density 50/50 Volume Split C) EB follower density, 60/40 Volume Split D) WB follower density 4 0/ 6 0 Volume Split E) EB follower density, 70/30 Volume Split F) WB follower density 3 0/ 70 Volume Split 0 5 10 15 20 25 30 35 0 500 1000 1500 2000 2500 3000 3500 4000 Follower density (veh/mi) Volume (veh/h) 50% no-passing-allowed and 50% passing-allowed 100% no-passing-allowed 100% passing-allowed 0 5 10 15 20 25 30 35 0 500 1000 1500 2000 2500 3000 3500 4000 Follower density (veh/mi) Volume (veh/h) 50% no-passing-allowed and 50% passingallowed 100% no-passingallowed 0 5 10 15 20 25 30 35 0 500 1000 1500 2000 2500 3000 3500 4000 Follower density (veh/mi) Volume (veh/h) 50% no-passing-allowed and 50% passing-allowed 100% no-passing-allowed 100% passing-allowed
101 D E F Figure 3 9 Continued 0 5 10 15 20 25 30 35 0 500 1000 1500 2000 2500 3000 3500 4000 Follower density (veh/mi) Volume (veh/h) 50% no-passing-allowed and 50% passing-allowed 100% no-passing-allowed 100% passing-allowed 0 5 10 15 20 25 30 35 0 500 1000 1500 2000 2500 3000 3500 4000 Follower density (veh/mi) Volume (veh/h) 50% no-passing-allowed and 50% passing-allowed 100% no-passing-allowed 100% passing-allowed 0 5 10 15 20 25 30 35 0 500 1000 1500 2000 2500 3000 3500 4000 Follower density (veh/mi) Volume (veh/h) 50% no-passing-allowed and 50% passing-allowed 100% no-passing-allowed 100% passing-allowed
102 A B C Figure 3 10 Directional average speed results for tests with a passing lane A) EB Avg Speed, 50/50 Volume Split B) WB Avg Speed, 50/50 Volume Split C) EB Avg Speed, 60/40 Volume Split D) WB Avg Speed, 4 0/ 6 0 Volume Split E) EB Avg Speed, 70/30 Volume Split F) WB Avg Speed, 3 0/ 70 Volume Split 40 42 44 46 48 50 52 54 56 58 60 62 64 0 500 1000 1500 2000 2500 3000 3500 4000 Average Travel Speed (mi/h) Volume (veh/h) passing-lane-added 100% no-passing-allowed 40 42 44 46 48 50 52 54 56 58 60 62 64 0 500 1000 1500 2000 2500 3000 3500 4000 Average Travel Speed (mi/h) Volume (veh/h) passing-lane-added 100% no-passing-allowed 40 42 44 46 48 50 52 54 56 58 60 62 64 0 500 1000 1500 2000 2500 3000 3500 4000 Average Travel Speed (mi/h) Volume (veh/h) passing-lane-added 100% no-passing-allowed
103 D E F Figure 3 10 Continued 40 42 44 46 48 50 52 54 56 58 60 62 64 0 500 1000 1500 2000 2500 3000 3500 4000 Average Travel Speed (mi/h) Volume (veh/h) passing-lane-added 100% no-passing-allowed 40 42 44 46 48 50 52 54 56 58 60 62 64 0 500 1000 1500 2000 2500 3000 3500 4000 Average Travel Speed (mi/h) Volume (veh/h) passing-lane-added 100% no-passing-allowed 40 42 44 46 48 50 52 54 56 58 60 62 64 0 500 1000 1500 2000 2500 3000 3500 4000 Average Travel Speed (mi/h) Volume (veh/h) passing-lane-added 100% no-passing-allowed
104 A B C Figure 3 11 Directional PTSF results for tests with a passing lane A) EB PTSF 50/50 Volume Split B) WB PTSF 50/50 Volume Split C) EB PTSF 60/40 Volume Split D) WB PTSF 4 0/ 6 0 Volume Split E) EB PTSF 70/30 Volume Split F) WB PTSF 3 0/ 70 Volume Split 0 10 20 30 40 50 60 70 80 90 100 0 500 1000 1500 2000 2500 3000 3500 4000 Percent Time Spent Following (%) Volume (veh/h) passing-lane-added 100% no-passing-allowed 0 10 20 30 40 50 60 70 80 90 100 0 500 1000 1500 2000 2500 3000 3500 4000 Percent Time Spent Following (%) Volume (veh/h) passing-lane-added 100% no-passing-allowed 0 10 20 30 40 50 60 70 80 90 100 0 500 1000 1500 2000 2500 3000 3500 4000 Percent Time Spent Following (%) Volume (veh/h) passing-lane-added 100% no-passing-allowed
105 D E F Figure 3 11. Continued 0 10 20 30 40 50 60 70 80 90 100 0 500 1000 1500 2000 2500 3000 3500 4000 Percent Time Spent Following (%) Volume (veh/h) passing-lane-added 100% no-passing-allowed 0 10 20 30 40 50 60 70 80 90 100 0 500 1000 1500 2000 2500 3000 3500 4000 Percent Time Spent Following (%) Volume (veh/h) passing-lane-added 100% no-passing-allowed 0 10 20 30 40 50 60 70 80 90 100 0 500 1000 1500 2000 2500 3000 3500 4000 Percent Time Spent Following (%) Volume (veh/h) passing-lane-added 100% no-passing-allowed
106 A B C Figure 3 12. Directional follower density results for tests with a passing lane A) EB follwer density, 50/50 Volume Split B) WB follwer density 50/50 Volume Split C) EB follwer density, 60/40 Volume Split D) WB follwer density 4 0/ 6 0 Volume Split E) EB follwer density, 70/30 Volume Split F) WB follwer density 3 0/ 70 Volume Split 0 5 10 15 20 25 30 35 40 0 500 1000 1500 2000 2500 3000 3500 4000 Follower density (veh/mi) Volume (veh/h) passing-lane-added 100% no-passing-allowed 0 5 10 15 20 25 30 35 40 0 500 1000 1500 2000 2500 3000 3500 4000 Follower density (veh/mi) Volume (veh/h) passing-lane-added 100% no-passing-allowed 0 5 10 15 20 25 30 35 0 500 1000 1500 2000 2500 3000 3500 4000 Follower density (veh/mi) Volume (veh/h) passing-lane-added 100% no-passing-allowed
107 D E F Figure 3 12. Continued 0 5 10 15 20 25 30 35 0 500 1000 1500 2000 2500 3000 3500 4000 Follower density (veh/mi) Volume (veh/h) passing-lane-added 100% no-passing-allowed 0 5 10 15 20 25 30 35 0 500 1000 1500 2000 2500 3000 3500 4000 Follower density (veh/mi) Volume (veh/h) passing-lane-added 100% no-passing-allowed 0 5 10 15 20 25 30 35 0 500 1000 1500 2000 2500 3000 3500 4000 Follower density (veh/mi) Volume (veh/h) passing-lane-added 100% no-passing-allowed
108 CHAPTER 4 DEVELOPMENT OF AN AN ALYTICAL METHODOLOGY FOR TWO LANE HIGHWAY FACILITY ANALYSIS The new simulation tool developed, as described in Chapter 3, was used to develop an analytical methodology for analyzing two lane highway facilities with various features, especially two lane highway segments with signalized intersections. This methodology is generally consistent with the HCM methodologies for basic two lane highway segments and signalized intersections. 4 1 Outline of Methodology T he new methodology used the Yu and Washburn study (2009) as the starting point T he ma in idea of this methodology is to divide a two lane highway facility into appropriate segments with uniform features then evaluate traffic operations on each segment by the unified performance measures, and finally aggregate performance measures over the entire facility and obtain LOS accordingly Therefore, one of the key issues in this study is to properly segment a two lane highway facility combined with signalized intersections. To achieve this objective, two primary tasks were completed as follows : Determine upstream effective length of signalized intersection The presence of a signalized intersection can significant ly impact two lane highway operations. Vehicles that are approach ing the signalized intersection will start to decelerate and prepare to stop if facing red, or decelerate to follow a discharging queue during green. This causes i nterruptions to traffic flow. In situations where a left turn bay is not present on the intersection approach, main line traffic may also be affected by turning vehicles even during green time. Therefore, it is essential to know the location where such
109 eff ects to main line traffic become significant. The distance between this critical location and the intersection approach stop bar is defined as upstream effective length. Determine downstream effective length of signalized intersection The presence of sign alized intersections may also change the downstream traffic flow characteristics of a signal. In addition, downstream traffic flow will be affected by the right turn and left turn vehicles from minor streets Traffic flow will generally return to its forme r state after some distance downstream of the signal. The location at which traffic Another key issue is to select appropriate service measure ( s ) for evaluating the traffic operations on two lane highways with va rious features. ATS PTSF and PFFS are employed in the current HCM methodology (TRB, 2010) for determining the level of service on basic two lane highway s However, as control delay serves as the service measu re for evaluating signalized intersections those three service measures cannot be applied to complex two lane highway facilities The potential service measure percent delay was examined for its efficacy of representing two lane highway operations which is also described in this Chapter. 4 2 Simulation Experiments In the Yu and Washburn (2009) study, CORSIM was used to develop the upstream ef fective length model while TWOPAS was employed to deriv e the downstream effective length model. As the capability of modeling two lane highway segments has now been incorporated into CORSIM, a complete two lane highway facility with a signalized intersection can be set up in CORSIM for data collection.
110 4.2.1 Te sting Facility The experimental design primarily focuses on two lane highway segments in combination with signalized intersections. Therefore, a two lane highway facility inclu ding an intersection is analyzed The testing facility is 8 mi long, on level te rrain, with no passing lane present, and with a sig nalized intersection at the 4 mi point. The free flow speed is 60 mi/h on the major two lane highway segment, and 45 mi/h on the intersecting road Passing is allowed along the facility except for the NETS IM links that are used for modeling the signalized intersection It should be noted that the presence of a left turn bay on the signalized intersection approach c ould result in some difference in upstream effective length, compared with the situation of a no left turn bay. When the through flow rate in the opposing direction is relatively high and no protected left turn phase is provided, the availability of a left turn bay can reduce the impedance to mainline flow caused by left turn vehicles that are wait ing for acceptable turning gaps. Assuming the storage of the left turn bay is sufficient fo r the left turn demand, the flow rate in the opposing direction will have little effect on the through movement in the analysis direction. Thus, it is not necessary to consider the opposing flow rate in the model for upstream effective length for the situation of a left turn bay on the intersection approach Based on such considerations, the models for estimating upstream effective length were developed separately for both conditions, with and without a left turn bay. According to the guidelines provided by the Delaware DOT (2009) defining no passing zone extension for departure legs at an intersection, 200 ft NETSIM links were used in modeling the signalized intersect ion for the main line. The minimum passing sight distance requirement used in passing models developed in Chapter 3 prevent passing
111 maneuvers happening near the intersection even with the only 200 ft long no passing markings upstream of the signal. The obs ervations of the simulation animations also confirmed that the design of 200 ft NETSIM links are reasonable. A 5 mi long lead up segment is included for obtaining well develope d platoons that will enter the facility where the data wer e collected. And in or der to prevent passing vehicles aborting their passes or expediting their passes due to the end of a passing zone, a 5 mi long follow up segment is used. The follow up segment also serve s as the lead up segment for the opposite direction. Passing is allowe d on both the lead up segments and the follow up segments. However, the outputs on these segments were excluded in the following analysis. 4.2.2 Simulation Scenarios It has been discussed and id entified in the Yu and Washburn study (2009) that the key factors that may affect the length of upstream effective length are peak volume, D factor, percentages of left turn and right turn movements, effective green time cycle length and availability of left turn bay. Therefore, the si mulation scenarios established in this study were developed based on the above contributing factors. Factor 1: Peak volume Peak hour traffic is primarily used in traffic analysis as it represents the worst case for traffic operations and highest capacity demands ( TRB 2010 ). In this study, 15 min peak flow rates were used in the simulation runs. Although two lane highways usually serve relatively low traffic demands, a wide range of flow rates are included in the simulation experiments to fully captur e the relationship between flow rate and upstream/downstream effective length. The minor street flow rate was set to half of the major street flow rate in all the experiments.
112 Factor 2: Heavy vehicle percentage Heavy vehicles such as trucks, buses and recreati onal vehicles are generally large and have lower performance in braking and accelerating, especially on grades. The existence of heavy vehicles in the traffic stream can significant ly impact traffic operations. In the passing models developed in Chapter 3 the size of the Therefore, different options for heavy vehicle percentages are preferred in the simulation experiments. C O RSIM currently employs nine vehicle types in FRESIM ( where the tw o lane highway modeling capability was incorporated ) four of which represent trucks : Type 3, Type 4, Type 5, and Type 6 (Table 4 1) As Type 6 trucks (i.e., double bottom trailer trucks) have much lower desired free flow speed s than other types of truck s, even on level terrain, only Type 3, Type 4 and Type 5 trucks were included in the experiments. It should be noted that the NETSIM component uses a different numbering system for vehicle types (Table 4 1 ) The vehicle type number of a specific vehicle on a two lane highway segment (FRESIM link) will change into the corresponding vehicle type number once it moves onto a NETSIM link and vice versa. Factor 3 : D factor The D factor represents the proportion of traffic traveling in the peak direction. When a left turn bay is absent in the analysis direction, the traffic in the opposing direction may restrict left turning opportunities in the analysis direction, and as a result, increase the possibility of upstream queuing. Given this consideration, the D fact or is expected to be considered in the regression model for estimating upstream effective length if no left turn bay is provided. Factor 4 : Percentages of left turn and right turn movements Different turning percentage options were applied in the simulati on experiments, as they are also an
113 important traffic characteristic. Even with the presence of a turn bay, the lane changing maneuver or decelerating maneuver executed by turning vehicles may also influence traffic through movement. In order to keep the c onservation of main line flow rate, the equivalent amount of turning flows from minor road were used to compensate for the flows left from main line at the intersection. Factor 5 : Signal timing For two lane highway facilities, it is preferred to keep main line traffic moving without interruption for as long as possible. Therefore, a fully actuated signal timing plan is more appropriate and efficient for a signalized intersection on two lane highways. Three two phase signal timing plans were implemented in the experiments to obtain different average effective gre en time and average cycle length, which were used for the regression model development. Factor 6 : Availability of left turn bay The models for upstream effective length estimation were developed separately for the situation with a left turn bay and the on e without a left turn bay. As discussed before, the D factor was added as a variable in the regression model for the situation without a left turn bay to capture the impacts on upstream queuing by opposing flow rate. Based on the combination of different i nputs, a total of 90 scenarios with a left turn bay and 216 scenarios without a left turn bay were established in CORSIM, and 10 iterations of each scenario were run. The simulation time period is 15 minutes for each run. The details of the experimental de sign are listed in Table s 4 2 to 4 4 It should be noted that in the scenarios without a left turn bay, the two way flow rate of 2000 veh/h (50/50 directional split) on the major road will result in continuous queue growth It was identified from the CORSI M experiments that growing queues occur when both
114 directional flow rates reach 900 veh/h. Therefore, in order to avoid unrealistic results, the highest flow rate option of 1000 veh/h for the analysis direction was eliminated from the experiments without a left turn bay. 4.3 Vehicle Trajectory Based Data Processing Procedure In the previous methodology developed by Yu and Washburn (2009), the variation of link based average speed was used to determine upstream/downstream effective length. For th e upstream ef fective length component a 3 mi long facility with a signalized intersection at the 1 mi location in CORSIM was used to produce the simulation data. Then, the upstream s ection was divided into 132 ft 8 long links for identifying the variation of average speed versus distance For the downstream effective length component the two lane highway simulation program TWOPAS was employed to obtain the simulation data. To replicate the traffic platooning characteristics downstream the si gnal, the TWOPAS input Entering Percent Following ( EPF ) was estimated ba sed on (2003) methodology Next, the variation of average speed was used to determine downstream effective length. This hybrid simulation approach was applied because no simulation tool that is able to model both signalized intersection and two lane highway segment existed at that time. In this study, using the two lane highway modeling capability now incorporated into CORSIM, the previous models were expected to be valid ated by using the new simulation program. And for more accurate results, vehicle trajectory information is desired. The rules for determining the vehicles that are affected by the presence of a 8 This distance was determined based on trying to segment to a short distance, in order to obtain the upstream effective length as accurate as possible. Meanwhile, this distance cannot be so short such that a vehicle could pass completely over a link in one time step (1 secon d).
115 signal were added to a tool named VTAPE ( Vehicle Trajectory An alysis for Performance Evaluation ). VTAPE can process CORSIM TS0 files, which contain time step data for each simulated vehicle. More detail about the rules implemented is included in the following sections. 4.3.1 Determine Upstream Effective Length In order to obtain the average result of upstream effective length in a certain scenario, it is important to find out where each affected vehicle begins to decelerate significantly due to the presence of a downstream intersection. Based on the vehicle traject ory information, two types of upstream affected vehicles were identified. Type 1 u pstream affected vehicle This type of affected vehicle is the first to stop at the stop bar It usually starts to decelerate for the yellow or red as it is getting close to the intersection. Normally, decelerating maneuvers under such a situation are completed within 2 to 4 seconds (Figure 4 1A for example). Type 2 u pstream affected vehicle This type of affected vehicle join s the queue at the stop bar. When there is already a queue at the stop bar, either stopped during red time or has not been discharged completely during green time, the upcoming vehicles will be aware of the existence of the queue and prepare to decelerate earlier than the type 1 upstream affected vehicle. affected vehicle usually starts to decelerate at no less than 4 ft/s 2 for several consecutive seconds (Figure 4 1B for example). Therefore, the criteri a (i.e. the position where a vehicle starts to be affected by the signal ) include two requirements: The vehicle must be no more than 2000 ft upstream of a signalized intersection. The threshold of 2000 ft was determined based on the longest upstream effective
116 le ngth of around 1700 ft from the current experiments. T he decelerating operations related to passing maneuvers are very unlikely to happen within this distance. The vehicle must decelerate at no less than 4 ft/s 2 at the current time step and the previous time step and then the position at the previous time step is recorded as the critical position. A number of observations of the simulation animations confirmed the effectiveness of th e criteri a In each scenario, the critical position for each upstream a ffected ve hicle was collected. T he average value was calculated from all of the individual vehicle values and used for the overall upstream effective length. 4.3.2 Determine Downstream Effective Length Determining the downstream effective length is largel y a function of the definition of when a vehicle that has departed from an upstream signal queue returns to normal two lane highway operation In the Yu and Washburn study (2009) the average speed difference (based on 100 ft long link s ) between the situat ion with an intersection and the situation without an intersection was used to determine downstream effective length. The location where the average speed difference falls below a certain threshold is considered as the end of the downstream effective lengt h. However in this study, the downstream effective length is determined from an individual ve hicle perspective for more accuracy Extensive research on the vehicle trajectory data indicates that a downstream vehicle can be considered as having left the do wnstream effective area when it travels stably. Two stable situations may exist: 1) If a vehicle is not in a following mode, it is considered as traveling stably when it reaches its own desired free flow speed and stops accelerating for several successive seconds; 2) If a vehicle is in a following mode, it should at least reach the desired free flow speed for driver type 1 and
117 stop accelerating for several successive seconds as an indicator for being stable. Two types of downstream affected vehicles were id entified according to the vehicle trajectory information. Type 1 downstream affected vehicle This type of affected vehic le is the one that has completely stopped for the red before travel ing downstream of the signal It could be a vehicle from upstream fl ow or a turning vehicle from the intersecting road This type of vehicle usually accelerates from zero speed when being discharged (Figure 4 1A for example). Type 2 downstream affected vehicle This type of affected vehicle may join a discharging queue at the intersection during the green, or turn from the intersecting road without stop ping at the intersection. In this case, the vehicle normally accelerates from a relatively lower speed after entering the downstream segment (Figure 4 1B for example). It should be noted that the maximum acceleration for Type 5 trucks in CORSIM is 1 ft/s 2 Therefore, when heavy vehicles are involved in a scenario, the average downstream effective length is extended The criteri a for locating the end of the downstream eff ective length for each downstream affected vehicle also include two requirements: The vehicle must be delayed by an flow speed when it first gets onto the downstream link, it is cons idered as an intersection delayed vehicle. Unaffected vehicles with driver types other than 1 normally travel at their own desired free flow speeds, or at flow speed when in a platoon led by a drive type 1 vehicle. All turning vehicles are considered as downstream affected vehicles, as they always flow speed for the turning movement, even for the vehicles with the most aggressive drive r type.
118 The vehicle must reach the free f low speed for driver type 1, and travel constantly or decelerate with in two consecutive seconds. Based on the examination of a considerable amount of vehicular data, it was found that a vehicle may not reach a stable level if it stops accelerating for just 1 second. And a vehicle that has become stable may sometimes decelerate a little bit to maintain a safe headway when in a platoon. Within any consecutive 3 seconds after being stable for a vehicle, if its acceleration is zero or negative in any 2 seconds, it may have a positive acceleration for the third second. Therefore, 2 seconds for checking a or not. The position at the first time step of the two consecutive time steps is recorded as the ending position of the downstream effective length. The ending position of the downstream effective length for each downstream affected vehicle was then determined based on the above criteria For each scenario, the average value was cal culated to obtain the overall downstream effective length. 4 4 Model Development The development of mathematical models for estimating upstream and downstream effective length was performed through regression analysis, and is discussed in this section. For each scenario, the upstream/downstream effective lengths for all the vehicles were averaged over 10 runs. The aggregate results were used in the regression model development. 4.4.1 Upstream Effective Length Model Development The regression model for estim ating upstream effective length wa s expected to be similar to the form developed by Yu and Washburn (2009), as follows : (4 1 ) where Y is the dependent variable, is the intercept lm k and q are parameters and X l and X m are the independent variables The model for estimating the upstream effective length of a signal when left turn bay is present was developed based on 90 scenarios as follows :
119 ( 4 2 ) where Len eff_up = upstream effective length of signalized intersection (f t) v d = flow rate i n the direction analyzed (pc/h) 9 Cycle = average cycle length (s) %LT = percentage of left turn vehicle s in the direction analyzed g = effective green time The statistical results are shown in Table 4 5 From Table 4 5 it can be seen that all the explanatory variables are statistically significant at a 95% confidence level Th e coefficient signs are logical; for example, increasing flow rate in the analysis direction or extending cycle length will increase the upstrea m effective length, i ncreasing the percentage of left turn vehicles or the effective green time will decrease the upstream effective length With an adjusted R squared value of 0.9798, it is indicated that 97.98% of the variance in the dependent variable ( upstream effective length) was explained by variations in the independent variables. The well behaved model residuals illustrated in Figure 4 2 confirm the goodness of fit of the model as well. The relationships between the explanatory variables and the response are generally consistent with the previous model. The model for determining upstream effective length under the situation of no left turn bay takes the D factor into consideration. However, the D factor is not directly used 9 In this study, the passenger car equivalent factors used to convert mixed vehicle flows into an equivalent passenger car flows are 2.2 for Type 3 trucks, 2.5 for Type 4 trucks, and 2.8 for Type 5 trucks.
120 in the model, but refle cted in the variable of opposing flow rate. The model was developed based on 216 scenarios as follows : ( 4 3 ) where Len eff_up = upstream effective length of signalized intersection (ft) v d = flow rate in the direction analyzed (pc/h) v o = flow rate in the opposing direction (pc/h) Cycle = average cycle length (s) %LT = percentage of left turn vehicles in the direction analyzed The statistical results are listed in Table 4 6 It can be seen from Table 4 6 that all the explanat ory variables are statistically significant at a 95% confidence level A new variable describing the opposing flow rate is included in this model It should be noted that effective green time is not included in this model. Intuitively, one would consider this variable to have a significant effect; however, since the actuated signal timing variable alone is sufficient to also account for the effect of green time. The analyst should be aware that if they are analyzing a situation where the signal timing is not optimized for the traffic flow conditions, the results from the upstream effective length model may not be completely accurate The coefficient sign of th e variable desc ribing opposing flow rate is reasonable, because increasing the opposing flow rate will reduce the turning opportunities for the left turn vehicles in the analysis direction, leading to the
121 increment in the upstream effective length when no left turn bay i s provided The coefficient sign s for the variables cycle length and percentage of left turn vehicles are contrary to the ones in the model for the situation with a left turn bay For the variable cycle length, a lthough a longer cycle length is usually con sidered as the cause of increased delay and queues, it may offer more turning possibilities for waiting left turn vehicles, and consequently allow the following through vehicles to pass through the intersection. This positive influence may offset the negat ive influence created by longer cycle lengths in other aspects. In this way, the inverse relationship between upstream effective length and average cycle length could be explained. For the variable percentage of left turn vehicles, it is logical that under the situation without a left turn bay, increasing the proportion of left turn vehicles will increas e the upstream effective length because left turn vehicles have to yield the opposing through traffic when no protected left turn phase is provided. With t he adjusted R squared value of 0. 9387 it is indicated that 93.87 % of the variance in the dependent variable (upstream effective length) was explained by variations in the independent variables. The well behaved model residuals illustrated in Figure 4 3 confirm the goodness of fit of the model as well. It should be noted that in some very rare cases where no vehicles turn left from the mainline (e.g., left turn is prohibited ) it is appro priate to apply the model for the with left turn bay situation to e stimate the upstream effective length even no left turn bay is provided b ecause it is no longer realistic to have a penalty for opposing flow in such situations to determine the upstream effective length.
122 A comparison was made between the two upstream eff ective length models (i.e., with and without a left turn bay) using the same combination of inputs. The results confirm the expectation that the no left turn bay scenario would always result in a longer upstream effective length than the with left turn bay scenario (Figure 4 4). 4.4.2 Downstream Effective Length Model Development The regression model for estimating downstream effective length is in the similar form described in Eq uation 4 1 Initially, models were developed separately for the scenarios with a left turn bay and the scenarios without a left turn bay, which is similar to the upstream effective length model development. However, the experiment results showed that an upstream left turn bay has little impact on the downstream effective length. The refore, the simulation data from both facility conditions were combined and used to develop one model for estimating downstream effective length. The model for estimating down stream effective length of signal was developed based on 306 scenarios as follow s : ( 4 4 ) where Len eff_ down = downstream effective length of signalized intersection (ft) V d = flow rate in the direction analyzed (veh/h) %HV = percentage of heavy vehicles Cycle = average cycle length (s)
123 %LT = percentage of left turn vehicles in the direction analyzed g = effective green time The statistical results are listed in Table 4 7 From Table 4 7 it can be seen that all the explanatory variables are statistically significant at a 95% confidence level With an adjus ted R squared value of 0.90 14 it is indicated that 90. 14 % of the variance in the dependent variable ( down stream effective length) was explained by variations in the independent variables. The well behaved model residuals illustrated in Figure 4 5 confir m the goodness of fit of the model as well. The relationships between the explanatory variables and the response are as expected. The positive relationship between downstream effective length and flow rate can be explained in that increasing flow rate resu lts in a longer average queue length upstream of the signal. The longer an upstream queue is, the more time is required for the queued vehicles to accelerate and disperse, which eventually leads to an increasing downstream effective length. A similar reaso n can also be used in explaining the positive relationship between cycle length and downstream effective length. The existence of heavy vehicles in the traffic stream is another reason for the increment of downstream effective length, because of the relati vely lower acceleration capability of heavy vehicle s In contrast, an increasing number of left turn vehicles can reduce downstream effective length. Within the current experiments, the turning flow rate from the intersecting roads are as the same as the mainline turning flow rate in order to keep the conservation of mainline flow However, as they are discharged during different phases, the average number of queued vehicles entering the mainline downstream segment actually decreases. Effective gre en time for the mainline through movement
124 also plays an inverse role in downstream effective length. Given a certain cycle length, a more effective green time means a reduc ed chance of queue build up during a cycle. The factor opposing flow rate was also e xamined as an explanatory variable, as initially it was thought that this may affect the acceleration and dispersion of upstream queues. However, as was the case for the left turn bay variable, this variable was found to have very little effect on the mode l results. Therefore, this variable was eliminated from the final model. 4. 5 Examination of the Efficacy of Percent Delay as an Appropriate Service Measure for Two Lane Highway Facilities Yu and Washburn (2009) proposed Percent Delay ( PD ) as the primary se rvice measure for a two lane highway with a signalized intersection. PD is believed to be able delay is the only service measure that determines the level of service of a signalized intersection, it is convenient to transfer control delay to total travel delay to determine the overall PD of a complex two lane highway facility. PD has been incorporated into CORSIM as an output for two lane highway links ( Table 3 4 ). In the current HCM (TRB, 2010) two lane highway methodology, two lane during driving. Class I two lane highways usually serve to connect major traffic generators (e.g., citie s, states), on which drivers expect to travel at high speeds and with more comfort. Thus, both ATS and PTSF are included in the Class I LOS criteria. Only PTSF is used in determining the LOS of Class II two lane highways (e.g., shorter intra city routes), on which high speed is not of great importance to drivers. Class III two lane highways is a new class in the HCM methodology. Two lane highways
125 traversing deve loped areas, or along scenic areas, are classified as Class III highways. Given that the speed limit typically reduces due to the relatively higher activity level within developed areas, the percent of free flow speed is more meaningful than the absolute a verage travel speed in the determination of LOS. Therefore, PFFS serves as the service measure for Class III two lane highways. Follower density has been identified as an effective service measure for the evaluations of two lane highway operations in previ ous studies: 1) follower density reflects the effects of flow rate, speed and percentage of followers ( Van as, 2006 ); 2) follower density reflects that flow rate is the major factor that affects the performance of highways ( A l Kaisy & Durbin, 2008 ); and 3) follower density has a wide range of variation ( Oreg on Department of Transportation, 2010 ). Hence it is worth considering follower density is also suitable for a two lane highway facility including signalized intersections. CORSIM can now also produce fol lower density as an output for two lane highway links. All the service measures discussed above were tested with the same two lane highway facility in section 4.2.1 (the one with a left turn bay). Only passenger cars were included in the tests, as well as all three signal timing plans (Table 4 4). In order to capture the performance of the facility over the entire range of flow rates, the capacity on the testing facility in the situation with each signal timing plan was estimated. It was found that the capa city ranges from 1408 veh/h to 1478 veh/h in the three different situations. Therefore, two flow rate levels of 1200 veh/h and 1400 veh/h were added to the existing flow rate options. Ten iterations of each experiment were run, and the average simulation r esults of the ten iterations were obtained and analyzed As the
126 volume split is 50/50 in all the experiments, only the eastbound r esults are presented As the service measure results only vary slightly among the three signal timing plans, given that all th e other inputs remain the same, only the results from the scenarios with the first signal timing plan are presented here. Generally, it is expected that the relationship between a service measure and flow rate is close to a linear relationship and has a wi de spectrum over the range of flow rate. A dashed straight line starting from the origin is added in each figure to help evaluate the nonlinearity of each relationship. ATS vs. flow rate. Figure 4 6 illustrates the relationship between directional ATS and directional volume. The results show a similar trend to the one from the basic two lane highway test results (Figure 3 7 A). The difference in the results between the signal timing plans is slight, until the flow rate gets close to the capacity. Signal timi ng plan III, which has the smallest maximum green time for the mainline, results in the lowest ATS at the highest flow rate. Although a strong inverse relationship exists between ATS and flow rate, ATS e of a signalized intersection, which is evaluated based on control delay (TRB, 2010). In addition, as speed limit reduction is usually applied in the vicinity of a signalized intersection, the absolute value of ATS has little meaning in evaluating such facilities. Therefore, ATS is not appropriate for assessing the performance of a two lane highway facility with signalized intersections. PTSF vs. flow rate. Figure 4 7 illustrates the relationship between directional PTSF and directional volume. The results are similar in trend to those from the basic
127 two lane highway test results (Figure 3 8 A). The difference in the results between the signal timing plans is not significant. Nevertheless, due to the limited passing o pportunities in the proximity of a signalized intersection, this passing related service area. Thus, PTSF does not suit the performance evaluation of a two lane hig hway facility with signalized intersections. PD vs. flow rate. Figure 4 8 illustrates the relationship between directional PD and directional volume. Similar to the ATS results, the difference in the PD results between the signal timing plans is slight unt il the flow rate approaches the capacity. Signal timing plan III, which has the smallest maximum green time for the mainline, leads to the greatest PD at the highest flow rate. As PD considers delays experienced both on basic two lane highway segments and in the influence area of an intersection, it is appropriate for the analysis of complex two lane highways. The PD grows slowly in the moderate to high flow rate range which is consistent with the variation of the ATS results. However, a PD value greater t han 18.3 (the peak PD value in Figure 4 8) for a two lane highway facility including signalized intersections does not necessarily mean LOS F. Given the same length of two lane highway, increasing the intersection density will increase the total travel delay, which as a result increases the PD result. For consistency with other LOS methodologies in the HCM, LOS F is not considered to be applicable until the demand exceed s the capacity. Follower density vs. flow rate Figure 4 9 illustrates the relationsh ip between directional follower density and directional volume. The results with different signal timing plans have little difference. As mentioned before, follower density reflects the
128 effects of flow rate, speed and percentage of followers. However, foll ower density might be more useful for uninterrupted two lane highway facilities in that it does not describe and it is not appropriate to be applied to a two lane highway f acility on which speed limit reduces in some areas. The relationship between follower density and flow rate is closest to a linear relationship according to Figure 4 9 This also confirms the good eligibility of follower density as a service measure for un interrupted two lane highways. Further study is desired on this service measure, including field data collection, simulation calibration, and the determination of LOS criteria. PFFS vs. flow rate Figure 4 10 illustrates the relationship between directional PFFS and directional volume. The PFFS flow curve shares the similar trend with the ATS flow curve (Figure 4 6 ) because the free flow speed in the current tests remains the same along the mainline. PFFS is t he service measure for Class III two lane highways in the HCM 2010. Although Class III highways represent two lane highways traversing developed areas, they are still considered uninterrupted highway facilities from the HCM perspective and different from t wo lane highway facilities that include signaliz ed intersections I t is possible to extend PFFS application in to the analysis of a two lane highway facility with a signalized intersection as the delay inc urred in the intersection influence area can be t ranslated into the overall average travel speed, based on which the overall PFFS can be determined. If the posted speed limit varies along a two lane highway facility with intersections, it is more appropriate to consider the weighted average PFFS based on different speed limits o ver the entire facility in that
129 case Therefore, extra calculation steps might be involved in an analysis procedure using PFFS as a service measure, which might make it somewhat more cumbersome to use than PD. In summary, PD is the most appropriate service measure among the five candidate measures presented here for a two lane highway facility with signalized intersections. Meanwhile, it has been confirmed that follower density can serve as an effective service measure for th e assessment of uninterrupted two lane highway operations. 4 6 Overall Evaluation Methodology The evaluation methodology uses the same idea as t he Yu and Washburn study (2009), as mentioned in section 4.1. The detailed evaluation procedure is described below, followed by the validation of the methodology. 4.6 .1 Evaluation Methodology The vehicle trajectory approach for determining upstream and downstream effective length is in fact consistent with the definition of control delay (Figure 4 1 1 ). Therefore, the delay experienced within upstream and downstream effective segments can be fully accounted for by control delay. In this sense, it is rational to divide a two lane highway facility with signalized intersections into basic two lane highway segments and intersection influence areas. The procedure for evaluating the traffic operations on a two lane highway with a signalized intersection can be divide d into seven steps, as follows : Step 1 Determine the upstream and downstream effective length of the signa lized intersection for obtaining the intersection influence area.
130 Step 2 Segment the facility according to the location of the intersection influence area. The areas other than the intersection influence area are treated as basic two lane highway segments Step 3 Determine the delay on each basic two lane highway segment by calculating the difference between actual travel time and free flow travel time. Step 4 Determine the control delay of the intersection. Step 5 Sum up the delays calculated in Step 3 and Step 4. Step 6 Determine the free flow travel time along the entire facility. Step 7 Determine the Percent Delay and the corresponding level of service The HCM 2010 analysis procedures are recommended to accomplish Step 3 and Step 4. 4. 6 .2 Validati on of the Methodology To validate the efficacy of this methodology, a two lane highway facility in combination with two signalized intersections was employed (Figure 4 1 2 ) in the tests with three different flow rate levels. The validation analysis procedur e is given in the steps below. Step 1 Given a facility, determine the upstream effective length and downstream effective length, and then divide the facility into appropriate segments. Step 2 Establish a testing facility in CORSIM based on the segments o btained in Step 1. Step 3 Run the simulation test to generate the overall P ercent D elay ( PD ) for the entire facility, and the control delay for each signalized intersection. Step 4 Establish a basic two lane highway facility in CORSIM, with the same inputs as the one in Step 2 except for the signalized intersections.
131 Step 5 Run the simulation test to obtain ATS on the basic two lane highway facility set up in Step 4. Step 6 Use t he control delays obtained from Step 3 and the ATS on the basic two lane facility from Step 5 to calculate the overall PD Step 7 Compare the calculated PD with the aggregate PD produced by CORSIM. From the comparison results in Table 4 8 it can be seen that the calculated PD is lower than the aggregate PD in every flow rate level. This difference was caused by the underestimation of control delay in CORSIM. In CORSIM, only the link immediately upstream of the signal has a nonzero value of control delay ; thus, the link immediately downstream of the signal always has a control delay of zero. C ontrol delay includes initial deceleration delay, queue move up time, stopped delay and the final acceleration delay ( TRB 2010). Therefore, control delay also occurs on the link immediately downstream of signal. If the control delay calculated in CORSIM is adjusted to an appropriate value to account for the downstream delay the aggregate PD gets very close to the calculated PD The comparison was also made between the aggregate PD from CORSIM output and the calculated PD based upon the HCM calculations for ATS and control delay The difference in the results is caused primarily by the underestimation of the average travel speed o n two lane highways in the HCM 2010 (Figure 4 1 3 10 ). If the average travel speed on a basic two lane highway segment calculated by the HCM 2010 is adjusted to an appropriate value, the aggregate PD becomes very close to the calculated PD 10 The HCM ATS results in Figure 4 1 3 were calcul ated by Equation 15 6 in the HCM (TRB, 2010) (refer to Equation 2 1 in this document).
132 4.6.3 Percent dela y based LOS criteria The valid thresholds for the PD based LOS criteria should satisfy two requirements: 1) for an analysis of a two lane highway facility without signals, analysts should obtain the same LOS if switching from the HCM service measures to PD ; 2) for an analysis of a two lane highway facility with a signal analysts will not obtain a better LOS compared with a no signal situation. To determine the thresholds for the PD based LOS criteria, the analytical methodology described in section 4.6.1 w as used to carry out two sets of calculations. One set of calculations is based on an 8 mi basic two lane highway facility, and the other set of calculations is based on an 8 mi two lane highway facility with a signalized intersection at the 4 mi point. Th e percent delay was calculated at seven different flow rate levels on both testing facilities. As in the current HCM two lane highway methodology (TRB, 2010), three different LOS criteria are used for three different two lane highway classes, and all t hree LOS criteria were employed as references to determine the PD thresholds. The procedure for determining the PD thresholds is summarized as follows: Step 1. Develop the relationship s between flow rate and the HCM service measures ( i.e., ATS PTSF and P FFS ) for the two lane highway facility without signals based on the HCM methodology. Step 2. Determine the corresponding flow rates at the LOS boundary values for each service measure. Step 3. Develop the relationship between flow rate and PD for the two lane highway facility without signals Step 4. Determine the boundary values for each PD based level of service based on the flow rate thresholds obtained from Step 2.
133 Figure s 4 1 4 to 4 1 7 illustrate the thresholds for the PD based LOS criteria based on the HCM two lane highway LOS criteria. The relationship between flow rate and PD for the two lane highway facility without a signal is also plotted in each figure to confirm that inserting a signal on to a two lane highway will not result in a be tter LOS. As the sets of PD thresholds obtained based on the HCM LOS criteria which are different for each two lane highway class were found to be different from each other the PD based LOS criteria are determined separately for each two lane highway class. It should be noted that both PTSF and ATS are employed as the service measure for a Class I two lane highway Therefore, two sets of PD thresholds based on the Class I PTSF criteria and the Class I ATS criteria w e re combined to determine one set of PD thresholds The details of the PD based LOS criteria are listed in Table 4 9 4 7 Additional Guidance for Facility Segmentation It is common in the field that the posted speed limit will be reduced in the vicinity of a signalized intersection on a two lane highw ay. Considering the impractica lity of speed limit reduction design as an input in a model, the guidance for facility segmentation regarding this factor is discussed in this section. 4 7 .1 Segmentation Guidance When speed limit reduction design is applied in the upstream segment of an intersection, two different situations may exist: Vehicles have to decelerate due to a speed limit reduction prior to entering the influence area of a signalized intersection Vehicl es enter the influence area of a signalized intersection, and are regulated by a speed limit sign in dicating the reduced limit The CORSIM modeling logic dictates that vehicles usually start to decelerate at 4 ft/s 2 to adjust their speeds under the reduce d speed l imit requirement. Therefore, using
134 VTAPE for determining the upstream effective length for each vehicle mentioned in section 4.3.1, it is difficult to distinguish the vehicles decelerating due to speed limit reduction from the vehicles deceleratin g due to the signal. However, it is still believed that the delay experienced in the influence area of a signalized intersection (combination of upstream effective segment and downstream effective segment) should only come from the control delay. Based on this understanding, given a two lane highway facility with a speed limit reduction upstream of the signal, the facility segmentation is performed as follows : Step 1 Calculate the upstream effective l ength based on given conditions. Step 2 Locate the posi tion where the speed limit drops. Step 3 Check if the position of the speed limit change is further upstream from the intersection than the position where the calculated upstream effective segment begins. If so, divide the unaffected segment into two basic two lane highway segments with different free flow speeds (Figure 4 1 8 A). Otherwise, the overlapped part will be considered as a part of the upstream effective segment without special treatment (Figure 4 1 8 B). 4 7 2 Example An example is presented for illustrating the evaluation procedure when speed limit reduction design is applied upstream of a signal. T he input data for a two lane highway facility that includes a signalized intersection has the followi ng characteristics: Directional flow rate = 800 pc/h (in both directions) PHF = 1.00 5% left turns
135 100% passing zones in both directions (except for the vicinity of the intersection) Level terrain 0% heavy vehicles 12 ft lane widths 6 ft shoulders 0 access points/mi (on basic two lane highway segments) 60 mi/h base free flow speed on most of the facility 50 mi/h free flow speed on the 800 ft link immediately upstream of signal in the analysis direction 8 mi segment length, the intersection is located in t he middle Left turn bay is provided Average cycle length (during 15 min analysis period) = 57.3 s Effective green time for major road (during 15 min analysis period) = 33.8 s Step 1 Determine upstream effective length. Upstream effective length is estimat ed using Eq uation 4 2 Then: Step 2 Locate the position where the speed limit drops Based on the inputs, the speed limit drops 800 ft in advance of the signalized intersection.
136 Step 3 Check if the speed limit drops prior to where the upstream effective segment begins Because 800 ft is greater than 440.3 ft, the speed limit drops prior to where the upstream effective segment begins. Step 4 Subdivide upstream segment. The 4 mi upstream segment is di vided into 3 segments with different features. The length of each segment is determined as follows : L eff_up : upstream effective segment L up_ 2 : basic two lane segment with free flow speed of 50 mi/h L up_1 : basic two lane segment with free flow speed of 60 mi/h Step 5 Determine downstream effective length. Downstream effective length is estimated using Eq uation 4 4 as follows :
137 Step 6 Subdivide downstream segment. The 4 mi downstream segment is divided into 2 segments with different features. The length of each segment is determined as follows : L eff_down : downstream effective segment L down 1 : basic two lane segment with free flow speed of 50 mi/h Step 7 Estimate ATS on basic two lane segment. The average travel speed on each basic two lane segment is estimated with the procedure in Chapter 15 of the HCM 2010 (TRB, 2010) The basic two lane highway segments with free flow speed of 60 mi/h: The basic two lane highway segments with free flow speed of 50 mi/h: Step 8 Estimate control delay. The control delay for the analysis approach is estimated with the p rocedure in Chapter 18 of the HCM 2010 (TRB, 2010) Step 9 Determine the delay on each segment and total delay along the facility. The basic two lane highway segment L up_1 : The basic two lane highway segment L up_ 2 :
138 The influence area of the signalized intersection is the combination of the upstream effective segment and downstream effective segment. The delay occurr ing in this area is accounted for by the control delay in Step 6. The basic two lane highway segment L down 1 : Therefore, the total delay can be calculated as: Step 10 Determine PD and the level of service (LOS) As the facility in this example is close to a Class III two lane highway, t he LOS can be determined by comparing the PD value with the PD criteria for Class III highways in Table 4 9 By applying the criteria, it is indicated that the analysis direction operates at LOS C 4 8 Comparison of the New Models to the Previous Models As the research approach used in the current study is different from the one used in t he Yu and Washburn study (2009), the differences between the current analytical methodology and the previous one are understandable from the following aspects.
139 Simulation tool The current methodology was developed based on the new version of CORSIM with the modeling capability of two lane highways incorporated. Thus, a two lane highway facility with various features (e .g., passing lane, signalized intersection, etc.) can be modeled in CORSIM as a whole. However, as such modeling capability was not available before, the previo us methodology was developed by a hybrid simulation approach. In t he Yu and Washburn study (2009 ), CORSIM was employed to model a two lane, two way road upstream of a signal, while TWOPAS was used to model a downstream two lane highway segment. Although the TWOPAS input parameter EPF can be used to represent the effects of a signalized intersection, the traffic flow simulated in TWOPAS was not related to the traffic flow simulated in CORSIM. Thus, some inconsistencies may be ignored. Moreover, the passing models employed in the new CORSIM are different from the ones applied in TWOPAS, which is also a possible source of the differences. Experimental design. The most different component in the current experimental design from the previous one is signal timing. As it is desired to keep major flows unimpeded as much as possible three fully actuated signal timing plans were used in the current experiments as opposed to pretimed control used in the previous experiments Fully actuated signal control is more responsive and efficient than pretimed signal control. But at the same time, different signal control types will definitely result in different simula tion data, which becomes a source of the differences. Another source of differences can be the facility design used to determine the down stream effective length in the previous study. The average travel speed at each data collection station on a two lane highway segment representing the sit uation without
140 a signal was compared with the one on a two lane highway segment representing the segment downstream of the signal. However, as no lead up segment (for o btaining normal platoons on the analysis two lane highway segment) was included in simulating the two lane highway segment without a signal, the average travel speed was close to the free flow speed at the beginning and l eveled off after some distance. Thi s issue may result in the overestimation of the downstream effective length. The differences may also come from the different centerline markings on the upstream link. The current experiments used a 4 mi long upstream segment with pass ing allowed all the w ay except in the vicinity of the signal. However, the previous study used a 1 mi long upstream segment with no passing allowed all the way down to the signal (because CORSIM did not have the ability to model passing maneuvers in the oncoming lane at that t ime) Although it was indicated that passing maneuvers rarely happen when vehicles approach an intersection, that distance is usually less than 2000 ft if passing is allowed. The length of the passing zone has impacts on the pattern of platoons, which can lead to significant differences between the two methodologies. Algorithms for determining upstream and downstream effective length The previous methodology takes the combination of upstream effective length and acceleration distance as the influence area of a signal. The downstream effective length begins after the acceleration distance. Speed variation (based on 132 ft link average speed in CORSIM and 100 ft link average speed in TWOPAS) is the criterion for determining the beginning of the upstream effec tive length and the end of the downstream effective length.
141 In the curre nt methodology, since vehicle trajectory data was used for the model development, the criteria for determining upstream and downstream effective length mainly focused on each individua consistent with the concept of control delay, the combination of upstream effective length and downstream effective length is considered as the influence area of a signal. The delay occurr ing within the inf luence area of a signal can then be fully represented by control delay.
142 Table 4 1 CORSIM truck types Performance description FRESIM vehicle type NETSIM vehicle type Single unit truck 3 2 Semi trailer truck with medium load 4 6 Semi trailer truck with full load 5 7 Double bottom trailer truck 6 8 Table 4 2 Experimental d esign for the scenarios w / a left turn bay Option 1 Option 2 Option 3 Option 4 Option 5 Flow rate (analysis direction) (veh/h) 200 400 600 800 1000 Heavy vehicles ( % ) 0 6 12 Signal timing plan Plan I Plan II Plan III Left turn vehicles (%) 5 10 Table 4 3 Experimental d esign for the scenarios w/o a left turn bay Option 1 Option 2 Option 3 Option 4 Flow r ate (analysis direction) (veh/h ) 200 400 600 800 D factor 0.5 0.6 0.7 Heavy vehicles ( % ) 0 6 12 Signal timing plan Plan I Plan II Plan III Left turn vehicles (%) 5 10 Table 4 4 Signal t iming p lans Major phase (permitted left turn phase) Minor phase (permitted left turn phase) Green max (s ) Green min (s ) Yellow (s ) All Red (s ) Green max (s ) Green min (s ) Yellow (s ) All Red (s ) Plan I 44 10 4 1 22 6 4 1 Plan II 38 10 4 1 19 6 4 1 Plan III 50 10 4 1 25 6 4 1
143 Table 4 5 Statistical results of the upstream effective length model w / a left turn bay Explanatory variables Parameter t value Adjusted R 2 Constant 266.6569 16.7318 0.9798 ( v d /100) 2 3.046 8 22.8773 Cycle 8.6256 10.471 1 ( v d /100 ) %LT 0.971 5 9.55 60 g 14.101 8 14.1122 Table 4 6 Statistical results of the upstream effective length model w/o a left turn bay Explanatory variables Parameter t value Adjusted R 2 Constant 412.0206 12.6299 0.9387 ( v d / 5 00) 3 57.9968 15.9946 ( v o / 5 00) 3 85.1575 19.9209 Cycle 3.6558 4.6047 [ ( v d / 5 00 ) %LT ] 3 0.0327 10.1975 Table 4 7 Statistical results of the downstream effective length model Explanatory variables Parameter t value Adjusted R 2 Constant 701.3438 24.2451 0.90 14 ( V /100) 51.0164 7.9484 %HV 42.3531 41.8964 Cycle 13.8329 3.9211 ( V /100) %LT 1.7009 5.0018 g 16.7603 2.7970
144 Table 4 8. Comparison results for validation Percent delay (%) Directional flow rate (pc/h) CORSIM output *Calculated result (CORSIM) **Calculated result (HCM) 200 5.20 4.56 10.00 600 11.74 10.79 22.10 1000 15.89 14.15 41.10 The results were calculated based on the average travel speeds on the basic two lane highway segments and the control delays produced by CORSIM ** The results were calculated based on the proposed methodology in section 4.6.1. Table 4 9 LOS criteria for two lane highway facilities based on PD Level of service PD (%) Class I Highways Class I I Highways Class I II Highways A 9 12 9.5 B > 9 1 4 > 12 1 6 > 9.5 2 1 .5 C > 14 20.5 > 16 23 > 21.5 36.5 D > 20.5 30 > 23 36.5 > 36.5 55.5 E > 30 > 36.5 > 55.5
145 A B Figure 4 1. Trajectories of intersection affected vehicles A) Type 1 upstream /downstream affected vehicle B) Type 2 upstream /downstream affected vehicle
14 6 Figure 4 2. Residual histogram of the upstream effective length model w/ a left turn bay Figure 4 3. Residual histogram of the upstream effective length model w/o a left turn bay
147 Figure 4 4. Comparison of predicted upstream effective length s using two estimation models Figure 4 5 Residual histogram of the down stream effective length model 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 Upstream effective length (ft) Directional flow rate (pc/h) No left-turn bay situation With left-turn bay situation
148 Figure 4 6 Directional speed flow results Figure 4 7 Directional PTSF flow results 45 48 51 54 57 60 0 200 400 600 800 1000 1200 1400 Average Travel Speed (mi/h) Directional flow rate (veh/h) 0 10 20 30 40 50 60 70 80 90 100 0 200 400 600 800 1000 1200 1400 Percent Time Spent Following (%) Directional flow rate (veh/h)
149 Figure 4 8 Directional PD flow results Figure 4 9 Directional follower density flow results 0 4 8 12 16 20 24 28 0 200 400 600 800 1000 1200 1400 Percent Delay (%) Directional flow rate (veh/h) 0 2 4 6 8 10 12 14 16 18 20 22 0 200 400 600 800 1000 1200 1400 Follower density (veh/mi) Directional flow rate (veh/h)
150 Figure 4 1 0 Directional PFFS flow results 74 76 78 80 82 84 86 88 90 92 94 96 98 100 0 200 400 600 800 1000 1200 1400 Percent of Free Flow Speed (%) Directional flow rate (veh/h)
151 A Figure 4 1 1 T rajectory of delayed vehicle on time space plane A) Delayed vehicle with a complete stop at intersection B) Delayed vehicle without a complete stop at intersection C ) Delayed vehicle with a complete stop at intersection and a relatively long upstream/downstream effective length
152 B C Figure 4 1 1 Continued
153 Figure 4 1 2 Illustration of a two lane highway facility with two signalized intersections Figure 4 13. Comparison of CORSIM ATS and HCM ATS 35 40 45 50 55 60 0 200 400 600 800 1000 1200 Average Travel Speed (mi/h) Directional flow rate (veh/h) CORSIM result HCM result
154 Figure 4 1 4 PD LOS criteria based on PTSF LOS criteria for Class I two lane highways 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 0 200 400 600 800 1000 1200 1400 Percent Delay (%) Percent Time Spent Following (%) Directional flow rate (veh/h) HCM PTSF (basic two-lane highway) Percent delay (basic two-lane highway) Percent delay (two-lane highway with one signal) LOS A LOS B LOS C LOS D LOS E
155 Figure 4 1 5 PD LOS criteria based on ATS LOS criteria for Class I two lane highways 0 10 20 30 40 50 60 70 80 90 100 0 5 10 15 20 25 30 35 40 45 50 55 60 65 0 200 400 600 800 1000 1200 1400 Percent Delay (%) Average Travel Speed (mi/h) Directional flow rate (veh/h) HCM ATS (basic two-lane highway) Percent delay (basic two-lane highway) Percent delay (two-lane highway with one signal) LOS A LOS B LOS C LOS D LOS E
156 Figure 4 1 6 PD LOS criteria based on PTSF LOS criteria for Class II two lane highways 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 0 200 400 600 800 1000 1200 1400 Percent Delay (%) Percent Time Spent Following (%) Directional flow rate (veh/h) HCM PTSF (basic two-lane highway) Percent delay (basic two-lane highway) Percent delay (two-lane highway with one signal) LOS A LOS B LOS C LOS D LOS E
157 Figure 4 1 7 PD LOS criteria based on PFFS LOS criteria for Class III two lane highways 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 0 200 400 600 800 1000 1200 1400 Percent Delay (%) Percent of Free Flow Speed (%) Directional flow rate (veh/h) HCM PFFS (basic two-lane highway) Percent delay (basic two-lane highway) Percent delay (two-lane highway with one signal) LOS A LOS B LOS C LOS D LOS E
158 A B Figure 4 1 8 Illustration of facility segmentation A) Speed limit drops before the starting point of u pstream effective length B) Speed limit drops after the starting point of upstream effective length
159 CHAPTER 5 SUMMARY AND RECOMMENDATIONS 5.1 Summary The work presented in this dissertation focuses on the development of a simulation program and analytical methodology for two lane highways. As the capability of modeling two lane highways has been incorporated into CORSIM the following two lane highway facilities can be modeled within CORSIM: Basic two lane highway segments with passing maneuvers (including passing one vehicle or multiple vehicles at a time) in the oncoming lane. Two lane highway segments with a passing lane. Two lane highway segments connecting to signalized intersections. The simulation program also allows users to modify newly added parameters for mo deling two lane highway facilities or operations. T he new performance measure outputs (e.g., PTSF follower density and PD ) and passing statistics outputs are available for different purposes of analysis. As the two lane highway modeling logic applied in the simulation program is reasonably consistent with theories and field observations from previous studies the modeling results are generally reasonable and consistent with expected traffic flow theory for tw o lane highways. I n the accomplishment of the simulation modeling capability built in CORSIM it allow s users to develop comprehensive traffic networks including a variety of facility types for different purposes of analysis. Based upon this new simulation capability in CORSIM, the previous methodology for two lane highway facility analysis developed by Yu and Washburn (2009) was updated. The new methodology retains the concept of facility segmentation from the
160 previous methodology, but was developed in a different way. First, the testing facility including both two-lane highway segm ents and a signalized intersection was established integrally in CORSIM, while the previous methodology used a hybrid simulation approach. Second, the algorithms to determine basic two-lane highway segments and intersection influence areas were developed based upon the information of individual vehicles, instead of aggr egate link performance that was used in developing the previous methodology. The service measure percent delay proposed in the previous methodology was examined and verified for its efficacy at dete rmining the level of service of a two-lane highway facility including signalized intersecti ons. A percent delay based LOS criteria derived from the simulation results of a se t of experiments is proposed. In addition, guidance for facility segmentation when speed lim it reductions are applied in the vicinity of a signalized intersection is provided. The application of the methodology can be extended to other complex two-lane highway facilities that include not only signal ized intersections, but also other features (e.g., passing lanes). It is also concluded that follower density, which reflects the levels of flow rate, speed, and percentage of followers, can be an efficient service measure for uninterrupted two-lane highway facilities, as the experiment results indicate that the relationship between follower and flow rate is closest to a linear relationship, compared with the performance measures of average travel speed, percent time-spent-following, and percent free-flow speed.
161 5.2 Recommendations Although t he modeling capability of two lane highways incorporated into CORSIM is based on existing traffic flow theory and field observations for two lane highways there are several potential studies that can be done to further refine and/or validate the current modeling logic as well as improve the capabilities, as follows: Refinement of desire to pass ( DTP ) and willingness to move over ( WTMO ) algorithms through direct traveler input (e.g., through focus groups). Incorporation of additional performance measures as proposed in the literature. Modification of the graphical network editor (TRAFED) to be able to accommodate editing of two lane highway networks. Collection and use of field data to calibrate and validate the two lane highway modeling logic. Another potential future study is to use field data to va lidate the new methodology for two lane highway facilities with signalized intersections. Due to lack of field data in this study, th e methodology is only validated within CORSIM. Input data, such as truck fleet composition, acceleration ability and desir ed free flow speed distribution may affect the upstream/downstream effective length models in the methodology Thus, adjustments might be necessary when applying the methodology in practice. It is also recommended that the algorithms to determine the infl uence area of a signalized intersection be applied in the determination of control delay in CORS IM, as they share the same conceptual definition CORSIM produces a nonlinear relationship between average travel speed and flow rate, whereas the HCM 2010 provides a linear relationship. This can lead to significant differences in simulation and HCM results for two lane highway segments.
162 Thus, it is recommended that field studies be done to further investigate the speed flow relationship for two lane highway s. The LOS criteria for two lane highways certainly need more investigation F ollower density is a promising service measure for uninterrupted two lane highway facilities because it reflects flow rate, speed and percentage of followers, which are the e ssen tial factors that determine the performance on a two lane highway facility, and it has a near linear relationship with flow rate indicated by simulation results. For two lane highway facilities with signalized intersections, it is proposed that follower de nsity can serve as a supplementary performance measure to percent delay which is the primary performance measure used to determine the level of service of such facilities.
163 LIST OF REFERENCES AASHTO. (2001). A Policy on Geome tric Design of Highways and Streets, Fourth Edition. Washington, D.C.: AASHTO (American Association of State Highway and Transportation Officials). Al Kaisy, A., & Durbin, C. (2008). Evaluating New Methodologies for Estimating Performance on Two Lane Highw ays. Canadian Journal of Civil Engineering 777 785. Al Kaisy, A., & Karjala, S. (2010). Car Following Interaction and the Definition of Free Moving Vehicles on Two Lane Rural Highways. Journal of Transportation Engineering 925 931. Al Kaisy, Ahmed; Freedman, Zachary;. (2010). Estimating Performance on Two Lane Highways: Case Study Validation of a New Methodology. Transportation Research Record, 2173 72 79. Brilon, W., & Weiser, F. (2006). Two Lane Rural Highways: The German Experience. Transportation Research Record, 1988 38 47. Carlson, P. J., Miles, J. D., & Johnson, P. K. (2006). Daytime High Speed Passing Maneuvers Observed on Rural Two Lane, Two Way Highways: Findi ngs and Implications. Transportation Research Record, 1961 9 15. Catbagan, J. L., & Nakamura, H. (2006). Evaluation of Performance Measures for Two Lane Expressways in Japan. Transportation Research Record, 1988 111 118. Catbagan, J. L., & Nakamura, H. ( 2008). Desired Speed Distributions on Two Lane Highways Under Various Conditions. Transportation Research Record, 2088 218 226. Catbagan, J. L., & Nakamura, H. (2009). Probability Based Follower Identification in Two Lane Highways. Transportation Research Board 88th Annual Meeting (pp. 09 0103). Washington, DC: Transportation Research Board. Cohen, S. (2002). Application of Car Following Systems to Queue Discharge Problem at Signalized Intersections. Transportation Research Record, 1802 205 213. DelDOT. ( 2009, May). DelDOT Manual on Uniform Traffic Control Devices (MUTCD) Part 3: Markings Frequently Asked Questions. Retrieved Dec 19, 2009, from Delware Department of Transportation: http://www.deldot.gov/information/business/drc/manuals/deldot_mutcd_part_3_faq .pdf Dey, P. P., Chandra, S., & Gangopadhyay, S. (2008). Simulation of Mixed Traffic Flow on Two Lane Roads. Journal of Transportation Engineering 36 1 369.
164 Dixon, M. P., Kyte, M., & Sarepalli, S. S. (2003, January). Effects of Upstream Signalized Intersections on Two Lane Highway Operations Retrieved 8 23, 2010, from http://www.ltrc.lsu .edu/TRB_82/TRB2003 002075.pdf Dixon, M. P., Sarepali, S., & Young, K. A. (2002). Field Evaluation of Highway Capacity Manual 2000 Analysis Procedures for Two Lane Highways. Transportation Research Record, 1802 125 132. Durbin, C. T. (2006). Traffic Perfo rmance on Two Lane, Two Way Highways: Examination of New Analytical Approaches. Master Thesis of Montana State University. Facility Analysis and Simulation Team, Transportation Development Division, Oregon Department of Transportation. (2010, December). Mo deling Performance Indicators on Two Lane Rural Highways: The Oregon Experience. Retrieved 9 19, 2011, from OREGON.gov: http://www.oregon.gov/ODOT/TD/TPAU/docs/A_Backgrou nd/ModelingPerf.pdf Farah, H., & Toledo, T. (2010). Passing behavior on two lane highways. Transportation Research Part F 355 364. FHWA. (2003). Manual on Uniform Traffic Control Devices for Streets and Highways (MUTCD). Washington, D.C.: Federal Highway Administration (FHWA). Glennon, J. (1988). New and Improved Model of Passing Sight Distance on Two Lane Highways. Transportation Research Record, 1195 132 137. Glennon, J. C. (1988). New and Improved Model of Passing Sight Distance on Two Lane Highways. T ransportation Research Record, 1195 132 137. Harwood, D. W., Gilmore, D. K., Richard, K. R., & Dunn, J. M. (2008). Passing Sight Distance Criteria. Washington, D.C.: Transportation Research Board. Harwood, D. W., May, A. D. Anderson, I. B. Leiman, L. & Archilla, A. R. (1999). Final Report, NCHRP Project 3 55(3). Midwest Research Institute, University of California, Berkeley. Harwood, D. W., May, A. D., Anderson, I. B., Leiman, L., & Archilla, A. R. (1999). CAPACITY AND QUALITY OF SERVICE OF TWO LANE HIGHWAYS. Berkeley, California, USA.: Midwest Research Institute, University of California Berkeley. Harwood, D. W., Potts, I. B., Bauer, K. M., Bonneson, J. A., & Elefteriadou, L. (2003). NCHRP Project 20 7 (160): Two Lane Road Analysis Methodolo gy in the Highway Capacity Manual. Washington, D.C.: Transportation Research Board, National Research Council.
165 Harwood, D., Gilmore, D.K., Richard, K.R., Dunn, J.M, & Sun, C. (2008). NCHRP Report 605: Passing Sight Distance Criteria. Washington, D.C: Trans portation Research Board, TRB, National Research Council. Hassan, Y., Easa, S. M., & Halim, A. O. Abd El. (1996). Passing Sight Distance on Two Lane Highways: Review and Revision. Transportation Research Part A, Vol. 30, No. 6 453 467. Hassan, Y., & Easa, S. M. (1996). Passing Sight Distance on Two Lane Highways: Review and Revision. Transportation Research Part A 453 467. Higgins, L. L., & Beesinger, J. (2006). Driver Perceptions of Risk: Potential Approaches to Improving Driver Safety. College Station, TX: Southwest Region University Transportation Center. ITT Corporation, S. D. (2007). Interactive Highway Safety Design Model (IHSDM) Traffic Analysis Module Engineer's Manual. New York: ITT Corporation. Khoury, J. E. (2007). Incorporating Uncertainty into the Estimation of the Passing Sight Distance Requirements. Computer Aided Civil and Infrastructure Engineering 347 357. Kim, J. (2006). A Capacity Estimation Method for Two lane, Two way Highways Using Simulation Modeling. Dissertation of the Pennsylvani a State University. Kim, J. T., Courage, K. G., Washburn, S. S., & Bonyani, G. (2003). Framework for Investigation of Level of Service Criteria and Thresholds on Rural Freeways. Transportation Research Record, 1852 239 245. Koorey, G. (2002). Assessment o f Rural Road Simulation Modelling Tools. IPENZ Transportation Group Technical Conference 2002. Rotorua, New Zealand. Leiman, L., Archilla, A. R., & May, D. A. (1998). NCHRP Project 3 55(3) Capacity and Quality of Service of Two Lane Highways: TWOPAS Model Improvements. Washington, D.C.: Transportation Research Board, National Research Council. Lemessi, M. (2001). AN SLX BASED MICROSIMULATION MODEL FOR A TWO LANE ROAD SECTION. Proceedings of the 2001 Winter Simulation Conference (pp. 1064 1071). Arlington, V A: Inst Elect & Electronic Engineers. McLean, J. R. (1989). Two Lane Highway Traffic Operations Theory and Practice. New York: Gordon and Breach Science Publishers. Morriss, J. L. (2005). Identification of Preferred Performance Measures For The Assessmen t of Level of Service on Two Lane Highways. Master's Thesis of the University of Florida.
166 Polus, A., & Cohen, M. (2009). Theoretical and Empirical Relationships for the Quality of Flow and for a New Level of Service on Two Lane Highways. Journal of Transpo rtation Engineering 380 385. Romana, M. G., & Prez, I. (2006). Measures of Effectiveness for Level of Service Assessment of Two Lane Roads: An Alternative Proposal Using a Threshold Speed. Transportation Research Record, 1988 56 62. TRB. (2010). Highway Capacity Manual 2010. Washington, D.C.: Transportation Research Board. Van as, C. (2006). The Development of An Analysis Method for The Determination of Level of Service of Two Lane Undivided Highways in South Africa. Pretoria, South African: South Africa n National Roads Agency Limited. Vanvalkenburg, G. W., & Michael, H. L. (1971). Criteria for No Passing Zones. Highway Research Record, 377 1 19. Warnshuis, P. (1967). Simulation of Two Way Traffic on An Isolated Two Lane Road. Transportation Research 75 83. Washburn, S. S., & Li, J. (2010). Development of a Simulation Program for Two Lane Highway Analysis. Gainesville, FL: Center for Multimodal Solutions for Congestion Mitigation, University of Florida. Weaver, G. D., & Glennon, J. C. (1972). Design and Striping for Safe Passing Operations. Highway Research Record, 388 36 39. Wooldridge, M. D., Messer, C. J., Heard, B. D., Raghupathy, S., Parham, A. H., Brewer, M. A., et al. (2001). Design Guidelines for Passing Lanes on Two lane Roadways (Super 2) College Station, Texas: Texas Transportation Institute, The Texas A&M University System. Yan, X., Radwan, E., Zhang, F., & Parker, J. C. (2008). Evaluation of Dynamic Passing Sight Distance Problem Using a Finite Element Model. Journal of Transportation E ngineering 225 235. Yu, Q., & Washburn, S. S. (2009). Operational Performance Assessment for Two Lane Highway Facilities. Journal of Transportation Engineering 197 205.
167 BIOGRAPHICAL SKETCH Jing Li was born in Dunhuang, China She studied traff ic and transportation After that, she entered the Graduate School of Beijing Jiaotong University in 2005 and Jing started her PhD program at the University of Florida in fall 2008, and became Research Assistant in Jan. 2009 under the supervision of Dr. Scott Washburn. During her PhD study, Jing worked on the projects two lane highway simulation and analysis. Jin g has been awarded primary research interest is traffic simulation and traffic operation analysis.