UFDC Home  myUFDC Home  Help 



Full Text  
DEVELOPMENT OF AN ARTERIAL LINK TRAVEL TIME MODEL WITH CONSIDERATION OF MIDBLOCK DELAYS By ALEXANDRA KONDYLI A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2005 Copyright 2005 by ALEXANDRA KONDYLI This document is dedicated to the graduate students of the University of Florida. ACKNOWLEDGMENTS The author would like to thank her graduate advisor, Dr. Lily Elefteriadou of the University of Florida for her insights and guidance throughout this thesis and her valuable support. The author wishes to thank the remaining members of the thesis committee, Dr. Scott Washburn and Dr. Ruth Steiner, for their assistance and their advices. Finally, the author expresses her sincere thanks to TransAssociates Consultant Firm at State College, PA, for their assistance during the data collection. TABLE OF CONTENTS A C K N O W L E D G M E N T S ................................................................................................. iv LIST OF TA BLE S ............................... ....... .. .. .. ........... .............. .. vii LIST OF FIGURES ............. ............. ........ ....... .......................... viii A B ST R A C T .......... ..... ...................................................................................... x CHAPTER 1 IN TR OD U CTION ............................................... .. ......................... .. B a c k g ro u n d ............................................................................................. 1 Problem Statem ent .................. ............................. .. ........... ............. .. O bjectiv e s ................................................................... ................................. . .2 2 LITER A TU R E REV IEW ............................................................. ....................... 5 H ighw ay C capacity M annual ........................................ ................................. 5 R ight T urns from A rterial ............................................................................. ..... .8 L eft Turns from A rterial ........................................... ....................................... 13 Access Management and Driveway Spacing...........................................................14 The Significance of MidBlock Effects ....................................... ...............16 Sum m ary of the Literature Review ................................. ................... ...... ........ 17 3 M E T H O D O L O G Y ........................................................................ .......................19 Data Collection .................. ............... ........................................... 19 Simulation Model Development and Calibration.....................................................19 D design of E xperim ents ....................................................................... ..................20 D database Expansion...................................................................... .. .......... 21 Data Analysis and Formulation of Regression Models............................................22 4 D A T A C O L L E C T IO N ................................................................... ....... ...............25 D ata R equirem ents......... .................................................................. .. ....... .... ..25 D description of Study Sites ................................................ .............................. 26 D ata C collection M ethods ........................................ .............................................27 v 5 SIMULATION MODEL DEVELOPMENT.............................................................31 S im u latio n P ack ag e .......................................................................... ........... .. ..3 1 M odel D evelopm ent .............................. .......................... .. ........ .... ..... ...... 32 M odel Calibration ......... .. ................ .. .... ........... ......... .. ... .... 35 Sum m ary and C onclu sions .............................................................. .....................39 6 ANALYTICAL MODEL DEVELOPMENT........................... ....................41 D database E expansion ............................ .. .... .................... 4 1 Selection of Simulation Output Performance Measures..............................44 D database O organization ......... ............................................................ .. ... .... ....... 45 Data Analysis....................................................... ...... ........ 46 Selection of Candidate V ariables.......................................... ........... ............... 46 R egression M odels........... ... ......... ...... ..................... .. .......... ..... ...... .. 54 Regression Model for TwoLane OneWay Arterials.................... ............... 59 Regression Model for TwoLane TwoWay Arterials .....................................63 DiscussionDescription of Independent Variables............... ......... ............... 68 C o n c lu sio n s..................................................... ................ 7 3 7 CONCLUSIONS AND RECOMMENDATIONS ............................................... 76 APPENDIX A PHASINGTIMING DIAGRAMS .............................................................. 81 B TURNING MOVEMENT AND LOOP DETECTOR DATA ..................................88 C TR A V E L TIM E STU D Y ................................................................ .....................97 D EX A M PLE PR O B LEM ............................................................................ ....... 100 L IST O F R E FE R E N C E S ........................................................................ ................... 102 BIOGRAPHICAL SKETCH ............................................................. ............... 104 LIST OF TABLES Table p 41 Determination of number of vehicle runs based on field measured travel time.......28 51 Calibration parameters for Park Avenue midday and p.m. models .......................37 52 Calibration parameters for Sparks p.m model ............... ................ ................... 37 53 Calibration parameters for Pugh midday and p.m. models.................................37 54 Field measured vs. simulation travel time after calibration ...................................38 61 Twolane twoway simulation model inputs for database expansion ....................43 62 Twolane oneway simulation model inputs for database expansion. First group of scenarios (from Beaver Avenue at Sparks Street) ............................................ 43 63 Twolane oneway simulation model inputs for database expansion. Second group of scenarios (from Beaver Avenue at Pugh Street)....................................44 64 Selected performance measures extracted from CORSIM ....................................45 65 Midblock delay equations for twolane oneway uncongested conditions.............60 66 Midblock delay equations for twolane oneway congested conditions ...............62 67 Midblock delay equations for twolane twoway uncongested conditions.............64 68 Midblock delay equations for twolane twoway congested conditions................66 LIST OF FIGURES Figure page 11 Illustration of midblock delay phenomena ..... ......... ..................................3 41 Twolane twoway arterial (Park Avenue)........ ..............................................26 42 Twolane oneway arterial (Beaver Avenue)................. ...... ... ........................ 27 43 First twolane oneway arterial link (Beaver Avenue between Sparks St and A th erto n S t) .................................................................... ................ 2 9 44 Second twolane oneway arterial link (Beaver Avenue between Pugh St and G a rn e r S t) ......................................................................... 2 9 45 Two successive twolane twoway arterial links (Park Avenue between N. Atherton St and N. Allen Rd and Park Avenue between N. Allen Rd and Shortlidge R d) ........................................................................30 61 Sketch of variable lVdr/N ................. ...................................... 48 62 Sketch of variable V a/N ......... .................................................... ............... 50 63 M easurement of discharge to demand ratio. ................................. .................54 64 Dataplot of midblock delay vs. discharge to demand ratio for twolane oneway arterials. .............................................................................56 65 Dataplot of midblock delay vs. discharge to demand ratio for twolane twoway arterials. (A) Midblock delay and discharge to demand ratio for low volume level. (B) Midblock delay and discharge to demand ratio for high volume level...57 67 Relationship between midblock delay and independent variables for the two lane oneway uncongested model. (A) Average arterial turning volume, IVan/N. (B) Average driveway turning volume, lVdr/N .............. ................... ................61 68 Relationship between midblock delay and independent variables for the two lane oneway congested model. (A) Average arterial turning volume, lVat/N. (B) Average driveway turning volume to demand ratio, (lVdr/D)/N. ...................63 69 Relationship between midblock delay and independent variables for the two lane twoway uncongested model. (A) Interaction between total arterial left turning volume and total arterial opposing volume, (lVartL* IVopp)/104. (B) Number of driveways per 1000 ft, Ndr ........................... ................................. 65 610 Relationship between midblock delay and independent variables for the two lane twoway congested model. (A) Arterial volume to demand ratio, Vup/Demand. (B) Interaction between total arterial leftturning volume and total arterial opposing volume, (IVartL* IVopp)/10 4.............................. ................. 67 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science DEVELOPMENT OF AN ARTERIAL LINK TRAVEL TIME MODEL WITH CONSIDERATION OF MIDBLOCK DELAYS By Alexandra Kondyli December 2005 Chair: Ageliki Elefteriadou Major Department: Civil and Coastal Engineering This thesis presents analytical models for estimating arterial travel time with consideration of delays at midblock locations. The midblock delays are defined as the delays that through drivers experience due to turning maneuvers of either the other major stream vehicles ahead that exit from the arterial or the minor stream vehicles that enter the arterial. These delays typically occur at the intersections with driveways. Volume and travel time data are collected in twolane twoway and twolane oneway arterials. These data are used for the development of simulation models and the expansion of the database through simulation. The generated data are used for the development of analytical equations of midblock delay through regression. The final regression equations provide estimation of arterial midblock delay depending on the conditions that the arterial operates (congested uncongested conditions). CHAPTER 1 INTRODUCTION Background Arterial roadways are designed to provide both accessibility and mobility to the users. Those two contradictory functions define the level of control of the arterials. There can be variable combinations of these, with respect to the land use and roadside development. According to the Highway Capacity Manual [HCM 2000] (Transportation Research Board, 2000) and the Policy on Geometric Design of Highways and Streets (AASHTO, 2001) arterials can be designated as highspeed, suburban, intermediate and urban based on their design and as principal or minor, based on their functionality. Arterials play a very important role in the roadway system. A quantitative assessment of the factors that affect midblock performance on urban arterial streets is important for the determination of the total link delay, as it is perceived by the drivers. The estimation of midblock arterial delay is important for the following reasons. First, it provides a complete delay estimation procedure that is partially and not wholly dependent on the intersection delay. Second, it is closer to reality, and it can assess the significance of the various midblock phenomena into the overall arterial link performance. Problem Statement Several traffic studies have focused on the operational characteristics of the arterial streets and their capacity, indicating that the estimation/prediction of delays and travel times is important information for the users. However, no study has developed a model for estimating travel time on arterial links by including various parameters of arterial midblock performance. Objectives The main objective of this research is to develop an analytical model that will predict arterial link travel time considering the delays of midblock phenomena. The mid block delays are defined as the delays that through drivers experience due to turning maneuvers of either the other major stream vehicles ahead that exit from the arterial or the minor stream vehicles that enter the arterial. These delays typically occur at the intersections with driveways. The model will also include delays due to parking activities and bus stops. Figure 1 1 illustrates the midblock phenomena that contribute to the increase of arterial travel time and that are explored in this research. The upper part of the figure shows a vehicle from the arterial (vehicle 1) reducing speed to make a right turn on driveway #1. This maneuver will force the oncoming vehicle to decelerate as well in order to maintain a safe distance. Similarly, vehicle 2 enters the arterial with lower speed than the oncoming vehicle. This will possibly cause the oncoming vehicle to decelerate. The lower part of Figure 1 1 illustrates an arterial vehicle (vehicle 3) about to perform a parking maneuver and by doing so, it will decelerate and cause the following vehicle to decelerate as well. Lastly, as vehicle 4 performs a leftturn maneuver from the driveway onto the arterial it can force oncoming vehicles from both directions to reduce their speeds to avoid a collision. Vehicle tunintg nght DRIVEWAY 2 nt the 1 OnconDRIVEWAYg eh3 DRIVEWAY 4 Ongoing vehcle nd .. .g nt. the dne, y DRIVEWAY 1 DRIVEWAY 3 DRIVEWAY 4 ffi~ Oncomn g vehicle decelerates Vehicle perform es parking maneuver I Vehicle turning left onto the arteal I __{_ I " n "7 " Oncomng vehicles fom both directions decelerate Figure 11 Illustration of midblock delay phenomena The tasks of this research are as follows: * Critically review all pertinent literature that involves arterial travel time estimation as a function of the arterial turning maneuvers. * Conduct field measurements of volumes and travel times in urban arterials to be used for the development and calibration of the simulation models. The study streets are twolane, twoway and twolane, oneway arterials. * Generate new data according to a prespecified design of experiments. * Use the expanded dataset for the development of analytical models using regression analysis. The final models express the midblock delay as a function of various parameters such as the arterial through and turning volume, the driveway volume, the arterial degree of saturation and others. The next chapter of this thesis summarizes the literature review. The third chapter presents the methodology to be followed in this research. The fourth chapter presents the data collection effort and the arterials under study. The fifth chapter discusses the simulation model development and the model calibration. The sixth chapter describes the experimental design and the development of the analytical models through regression. r F 4 The last chapter discusses the findings of this thesis and presents recommendations for further research. CHAPTER 2 LITERATURE REVIEW The literature review includes several components. First, the respective chapters of the HCM 2000 have been reviewed. Additionally, studies that estimate delays due to right or left turn maneuvers are presented. Also, studies that report the effect of driveway spacing as a result of access management techniques to the overall arterial operations are also reviewed. Finally, literature that suggests that arterial delay models should incorporate the effects of midblock phenomena is presented. Highway Capacity Manual The HCM 2000 provides methodologies for the evaluation of urban streets by determining arterial link Level of Service and computing the intersection delay. Information related to the urban streets methodology can be found in Chapter 15. The first step of the methodology of urban streets is to define the arterial's classification, which is based on both the design and functional characteristics. The design characteristics of the arterials are related to the arterial posted speed limit, the signal density, the driveway/ access point density and other design features. Based on these characteristics, the arterial street classification is determined, which, in the next step, affects the arterial running speed, along with other parameters, such as the freeflow speed, and length of the segment. The arterial running speed is very important for the LOS analysis, because when combined with the intersection control delay, it is used for the calculation of the arterial throughvehicle travel speed for the segment or for the entire link under consideration. According to the travel speed and the arterial classification, the level of service of the arterial segment is calculated. Evaluating the arterial analysis methodology of the HCM 2000 (Chapter 15) it is concluded that it does not explicitly account for phenomena such as driveway density, crossstreet traffic blocking the through movement, and arterial turning maneuvers that impede the oncoming vehicles. This chapter mentions that the arterial running time is affected by the presence of parking; however the methodology does not quantify this effect. The HCM 2000 also includes methods for estimating delays at both signalized and unsignalized intersections. These methods are widely used and are accepted among engineers. For signalized intersections, the HCM 2000 (Chapter 16) provides a procedure for estimating the delay and the level of service. The methodology calculates the control delay, which includes movements at lower speeds and stops at intersection approaches as vehicles move up in a queue or slow down at an intersection. The delay formula accounts for uniform arrivals, for random arrivals and oversaturated queues and for initial queues. The formula that estimates the delay assuming uniform arrivals is based on the Webster's formula and it is widely accepted and used in practice. The presence of buses and parking frequency is taken into consideration for the saturation flow rate methodology (HCM 2000, Chapter 16). For the determination of the saturation flow rate two adjustment factors are introduced; the busblockage adjustment factor and the adjustment factor for existence of a parking lane and parking activity adjacent to the lane groups. However, these factors are considered to affect the traffic stream only within 250 ft from the signal and not at midblock locations. For twoway stopcontrolled intersections and Tintersections with a single minor street approach, the HCM 2000 (Chapter 17) provides methods for estimating delays and levels of service for the minor approaches caused by the priority approaches. The delays are calculated for minor approach vehicles that are crossing the major street, turning right on the major street, turning left on the major street, and turning left from the major street, depending on the minor street volume, major street volume, and followup time. Delays are also calculated for the major street through or right turn vehicles that are impeded by the left turning vehicles when there is a shared lane on the majorstreet approach and no exclusive leftturn pockets are provided. The HCM 2000 also notes that these delays usually 'have very small effect because the major street usually provides enough space for the blocked (through) vehicle to sneak by or bypass the leftturning vehicle'. Based on the HCM 2000, it can be concluded that a more detailed analysis of the segment (on the level of individual majorminor street intersection) would require the application of the unsignalized intersections methodology, which, according to the aforementioned, provides the delays of the minor approaches and only of the major approach left turn. The HCM 2000 does not provide a methodology for calculating the delays that the major street vehicles may incur due to rightturn maneuvers from the arterial, rightturn maneuvers and left turn maneuvers from the minor street, and such delays are not taken into consideration for the analysis of the entire arterial segment. More specifically, the HCM 2000 mentions that ". .. in special cases, there might be midblock delays caused by vehicle stops at pedestrian crosswalks, or other delays caused by bus stops or driveways." These midblock delays can be directly incorporated into the methodology provided that the user already has an estimate of their value. Right Turns from Arterial Two studies that were conducted in 1970 are focused on the impact of rightturn vehicles to the delay of the through vehicles. In the first study, Stover et al. (1970) used simulation to quantify the effect of rightturning vehicles. For the calibration of the model, deceleration and rightturn speed data from aerial timelapse photographs were used. The simulation analysis considered the effect of majorstreet flow rate, proportion of rightturns, and driveway entrance speed. The authors' simulation results show that through vehicle delays increased with increasing majorstreet flow rate, and are higher under low rightturning speed, but their findings may not be valid nowadays due to changes in the driving behavior. The second study was conducted by Alexander (1970). He observed traffic operations at seven mostly urban intersections on twolane twoway highways in Indiana to determine the delay to through traffic due to rightturning vehicles. The following equation was developed based on a regression analysis of the fieldmeasured delay and flow rates: Dt = 219 + 2.OSQr + 0.37Q + 4.33u (2 1) Where: Dt = total through vehicle delay, Qr = rightturn flow rate, vph, Q = majorstreet flow rate, vph, u = majorstreet running speed, m/s. The R2 for this equation was reported to be 0.76. The author's findings are that the total delay caused by the rightturn vehicles is related to the volume of the rightturn maneuver, the volume and the speed of the through vehicles. McShane (1995) used the TRAF/NETSIM (1995) simulation model to quantify the effects of rightturn maneuvers on through vehicle travel speed. The author took into consideration the driveway flow rate, driveway spacing, majorstreet flow rate, driveway location, number of driveways, number of lanes on the major street, and freeflow speed. The computed delays are comparable to those reported by Stover et al. (1970) and Alexander (1970), although, an exact comparison is not possible due to different ranges of majorstreet flow rate and speed. Bonneson (1998) developed a deterministic model for predicting the delays to majorstreet through drivers due to a rightturn maneuver from the outside through lane of the major street. The author did not consider the number of through lanes on the major street or the distribution of its flow rate to these lanes. The proposed model requires as input the flow rate in the outside through lane and it is formed for both single and multilane approaches where rightturns are assumed to occur from the outside through traffic lane. The author modeled the delay of the through vehicles that starts with the rightturn maneuver of one vehicle and ends with another rightturn maneuver. In this model it is assumed that lane changing by through drivers to avoid a slowing rightturn vehicle is negligible during the event, due to the fact that the event has relatively short duration and the delays are basically because of the acceleration/deceleration process (only a few seconds). The model describes first the delay incurred by the first through vehicle and then the delay of the following vehicles by representing the trajectories of the turning vehicle and the through vehicles, under the assumption of low flow conditions (1000 vph/ln), constant running speeds and constant acceleration/deceleration rates. Bonneson's desired turn speed was related to the curb return radius and the driveway width, based on research by Richards (reported by Stover and Koepke, 1988). The formula used is rt = 3.59 + 0.196 R (2 2) where urt is the rightturn speed and Rc is the turn radius. The author also determined the minimum speed and the delay of the first delayed through vehicle. Based on shock wave theory, the author developed a procedure for calculating the delay of the following vehicles. The verification of the model entailed comparison of the proposed model with the findings of other researchers (Alexander, 1970; Stover et al., 1970; McShane, 1995) which yielded overall agreement, as well as a comparison of it to the TRAF/NETSIM model, but no validation with field data was conducted. The author's findings indicate that the through vehicles' delay increases with increasing flow rate, increasing major street running speed, with an increase in rightturn vehicles proportion or a decrease on rightturn speed. It is also shown that the delay per rightturn vehicle decreases as the proportion of the rightturn vehicles increases, due to the fact that the proportion of through vehicles that are following is smaller. In NCHRP Report 420, Gluck et al. (1999) analyzed types of access management techniques and their impacts. On the assessment of unsignalized access spacing, the authors performed an operational analysis for identifying how right turns entering a driveway affect other drivers following in the same travel lane. These findings are also shown in a paper review of Gluck et al. (2000). In their study, information was gathered on the number and percentage of through vehicles impacted by right turns. The impact lengths of through vehicles impacted were determined, and, influence areas were computed. Their results were used to quantify the effects of multiple driveways and to develop inputs for establishing unsignalized access spacing guidelines. The field measurements include traffic volumes and impact characteristics such as the number of incidents caused by the activation of the brake lights and evasive maneuvers of through vehicles following a rightturning vehicle. Moreover, the authors gathered information and computed the following input parameters: * The number and percentage of through vehicles in the right lane that were impacted by rightturnin at a single driveway. * The percentage of through vehicles in the right lane that were impacted by right turn in over a series of driveways. * The distances back from a single driveway entrance that cars began to be impactedthe impact lengthand the spatial distributions of impacted vehicles. * The "influence areas" or influence distances before (upstream of) a driveway entrance. This involved adding perceptionreaction distance and car length to the impact length. * The proportions of through vehicles in the right lane whose influence lengths extended to or beyond at least one upstream driveway over a section of road (spillback rate). * The variations of spillback rate by roadway operating speed The single driveway analysis was extended to multiple driveways' analysis through probability analysis. The findings denote a linear relationship between the percent of right lane through vehicles impacted and the rightturn volume, irrespective of speeds. The driveway impact lengths (and influence lengths) analysis revealed the relationship between the percentage of through vehicles in the right lane that would be impacted by rightturn traffic for various distances form a driveway for each range of rightturn volume. The analysis also revealed that the influence distance increases as speed increases. A relationship between speed, distance from upstream traffic signal and impact length is also established. Another study presented by Wolfe and Piro (2003) describes a methodology for determining the delay to through vehicles due to the right turning traffic. The study involves both signalized and unsignalized intersections, but the authors eliminated from their study the vehicles that were not under the operating speed due to being in a queue or decelerating at the amber traffic light. The methodology is based on total volume, right lane volume, right turning volume and the difference in the through vehicle operating speed and the right turning speed due to geometric constraints. The methodologies that were developed calculated three different forms of delay, such as the total delay of all through vehicles, the delay to traffic in the right lane and the delay to all through vehicles that follow a right turning vehicle. The following delay equation was derived based on the total volume, the right lane flow, the rightturning volume, and the algebraic difference between the rightturning vehicle speed and the through operating speed. Dtot = 0.352,tol + 0.729VRLne + 4.99VR, 5.97u R (2 3) The rightturning speed was calculated by using an equation of a previous study (Wolfe and Lane, 1999) that correlates the turning speed with the turning radius of the curb. The authors collected data at 15 intersections, either signalized or unsignalized, with various right curb radii, ranging from 1.6 m to 20 m. The authors determined the speed of the turning vehicles based on the time they required for the maneuver and the respective geometry of the intersection. Based on their data, the following formula was derived. S= 2.2678809 + 0.451631R + 0.078901R2 0.007308R3 + 0.0001811R4 (2 4) Based on that speed, the authors determined the time difference between the right turning vehicles that enter the arterial and the through vehicles, which holds for the delay that the through vehicles experience. Left Turns from Arterial Bonneson and Fitts (1999) discussed the delay on the major street due to vehicles that perform a leftturn maneuver at twoway stopcontrolled intersections. This delay is incurred when major street leftturn demand exceeds the available storage area and blocks the adjacent through lane (undivided cross section with no leftturn bay). In this situation, the through drivers will merge with vehicles in the adjacent through lane if there is an adequate gap for them to safely merge into, or they will remain in the inside lane until the queue dissipates (if there is no merge opportunity). This paper is part of the NCHRP Project Report 395 (Bonneson and McCoy, 1997) that evaluates the adequacy of midblock leftturn treatments such as TWLTL, raisedcurb median and undivided cross section, based on the operational, safety and access effects of these treatments. In this paper the authors combined four models for determining the average delay to through vehicles in a blocked inside lane. The four models include: a lane flow rate model, a merge capacity model, a merge delay model, and an overflow probability model. The authors also discuss the work of Kyte et al., (1996) who proposed a model to estimate the delays to through vehicles by assuming equal distribution of through traffic to lanes and the probability of having a leftturn queue. The lane flow rate model was developed to predict the through vehicle flow rate in each approach lane just upstream of the leftturn location, when there is at least one left turn vehicle present (for two or more lane approaches). The distribution of through drivers on the lanes follow the assumption that they will choose the traffic lane that minimizes their travel time (and thus their delay), and this is accomplished by equating the demandtosaturation flow ratios among the alternative through lanes. Additionally, the authors developed capacity models for the inside lane through vehicles, for either nonmerge or merge situations. The nonmerge situations occur when a driver decides to remain to the inside lane until the queue ahead of him dissipates and does not perform a lane change maneuver. The merge situations take place when the driver merges into the adjacent through lane, instead of waiting in the back of the queue. Lastly, the probability of a leftturn bay overflow is calculated, which represents the probability of one or more leftturn vehicles being queued in the inside through lane in an undivided cross section. The combination of the models yields that the delay of the through vehicles increases with leftturn percentage for low to moderate flow rates; however, for high flow rates this is not the case. The authors' belief is that there may be a leftturn percentage associated with the maximum delay for high flow rates such that leftturn percentages higher or lower would yield lower delay. This methodology was verified using the TWLTLSIM simulation model and it was found that the two models generally agree, although the proposed model predicts lower delays than the TWLTLSIM in the 'low delay' range. Access Management and Driveway Spacing According to the Access Management Manual, published by the TRB (2003) access management is "the systematic control of the location, spacing, design, and operation of driveways, median openings, interchanges, and street connections to roadway". In other words, access management is a tool for providing vehicle access to the abutting land development, in a way that the traffic safety and transportation efficiency are met in balance. There are several studies that developed guidelines for selecting the desirable spacing between unsignalized access points (Stover and Koepke 2002, AASHTO 2001, Gluck et al. 1999, TRB 2003). These guidelines were based on different criteria, such as safety, stopping sight distance, intersection sight distance, functional area, rightturn conflict overlap, influence distance and egress capacity (TRB, 2003). Based on a study by S&K Transportation Consultants Inc. (2000), as appears in Access Management Manual (TRB, 2003), the relative crash rates are expected to increase if the driveways' spacing is reduced. For example, a decrease of access spacing from 1056 ft to 264 ft would yield crash rates 2.1 times higher. Another study (Stover and Koepke, 2002) suggests that long spacing between driveways is more desirable, since auxiliary lanes can be designed to reduce the conflicts between the arterial through vehicles and the turning vehicles and provide safety. The AASHTO Green Book (2001) provides suggestions for stopping sight distance and intersection sight distance, which can be applied as access spacing criteria. Stopping sight distance is the minimum sight distance required to allow drivers to come to a stop. This criterion is very useful for access spacing guidelines, as there are many potential conflicts between the arterial through and turning drivers. AASHTO (2001) provides tables that give the minimum stopping sight distance, depending on the arterial grade. Additionally, intersection sight distance is the minimum sight distance required for a vehicle stopped at an intersection to enter or cross the major approach. AASHTO (2001) also provides minimum intersection sight distances as a function of the major street speed, number of lanes and grade. The functional area criterion (AASHTO, 2001) demonstrates that all access connections such as intersections (signalized or unsignalized), driveways and arterials, define a functional area upstream of their location. The criterion dictates that no other connection should be placed within this functional area. The rightturn conflict overlap occurs when the major street through driver has to monitor more than one driveway at a given time. The speed difference between the turning vehicle and the through vehicle define the minimum distance required to reduce collision due to overlapping rightturn maneuvers (Stover and Koepke, 2002). The Significance of MidBlock Effects Lin et al., (2003) developed a model which includes the effect of the vehicles entering an arterial from a cross street. According to the authors, the total delay on an arterial includes Link Delay and Intersection Delay. Link Delay can be caused by two factors. The first factor is the intersection delay, in the sense that a vehicle will slow down while approaching the queue at the intersection, hoping that the queue will start to dissipate as soon as he will reach the intersection. In essence, this represents an early deceleration, which does not affect delay. The second factor is the increase in flow, which can be questionable, because it has been shown that travel time is not sensitive to link flow, for medium or high flows, due to the metering of the upstream intersections. For this reason, the authors assume that the delay of the link that is due to the internal link flow can be considered zero. Under this assumption, the authors approximate the Link Delay with the estimation of only the delay at intersections. The model is based on discrete Markov chain properties. The key parameters of the model are the ratio of the overall flow level to the service capacity for the intersection in question, the net turning movement percentages into the arterial from the cross street at the upstream intersection, and the traffic signal coordination level with its upstream intersection. As it is seen, the authors acknowledge that the vehicles that turn into the arterial can produce delays to the through vehicles. The model that is developed, exhibits some desirable properties in predicting the arrival time at downstream arterial links, but it has some limitations and it has not yet been validated with field data. Olszewski (2000) compared the HCM 2000 methodology for estimating intersection delay and another speedflow model that requires intersection spacing and minimum signal delay as input parameters. This model was developed primarily for planning applications. The author compared the travel speeds that are predicted by both models, for a range of parameters such as intersection spacing, traffic flow and signal characteristics. The results yielded that the pattern of travel speeds is similar to both models; eventhough the HCM 2000 model predicts lower speeds. Summary of the Literature Review Several research efforts have been reviewed to establish the current stateoftheart that relates the arterial travel time and delay estimation with midblock effects. Most of the literature is devoted to access control management and evaluation of alternative median treatments. Other studies try to quantify the effects of either right turns or left turns from the arterials to the travel time of the through vehicles, by developing analytical models. Moreover, studies were found to acknowledge the significance of the midblock effects but without proposing any delay model that incorporates these effects. Thus, it can 18 be concluded that a comprehensive model that incorporates all the parameters that can cause arterial link delay to the through vehicles is lacking from the current literature. CHAPTER 3 METHODOLOGY This chapter presents the methodology followed to develop the analytical models of midblock delay of arterial streets. The following tasks were undertaken: 1. Collect volume and travel time data for twolane, twoway and twolane, oneway arterials. 2. Simulate the arterials in CORSIM and calibrate the models. 3. Conduct a factorial design experiment. 4. Expand the database and perform multiple simulation runs. 5. Extract the midblock delay data and develop midblock delay equations as a function of the independent variables of interest using regression analysis. A more detailed description of the process followed by this research is presented in the remaining of this chapter. Data Collection Data were collected at two twolane, oneway arterials and one twolane, twoway arterial. The data involve arterial through and turning volumes, driveway turning volumes, heavy vehicles, parking frequency, and bus dwell time, as well as travel time measurements. These data were used for the next step of the research (simulation model development and calibration). A more detailed description of the data collection process and the study sites is presented in Chapter 4 of this thesis. Simulation Model Development and Calibration One of the main purposes of this study is to use reallife traffic conditions for the development of the regression models. However, this would require a large amount of field measurements, which is outside of the scope of this research. Collecting a certain amount of field data and using a simulation tool for expanding the dataset would reach the research objectives. The study arterials were simulated in CORSIM and calibrated with the travel time measurements. The criterion used for deciding whether the models need to be calibrated or not is that the simulation travel time should be within a 10% range of the actual field measured travel time. The calibration of the models includes adjustments of the simulation parameters such that the resulting travel time would approach the field measured travel time and the reallife conditions would be replicated effectively. The calibration parameters considered are the discharge headway, the mean startup delay, the driver's reaction time, the driver familiarity, the probability of spillback, the duration of lane change maneuver, the parking maneuver duration, and the freeflow speed. Details regarding the simulation model development and calibration are given in Chapter 5. Design of Experiments Once the simulation models are created and calibrated, they were used as a basis for generating more data and expanding the database in CORSIM. The database expansion process requires the development of a factorial design in which several simulation model inputs are identified as varying factors with different levels, to attain variability in traffic and geometric conditions. The varying combinations between the levels of the simulation input parameters form different scenarios. The simulation input parameters were selected based on the arterial type, depending on whether it is a twolane, twoway or a twolane, oneway arterial. The simulation input parameters for the twolane, twoway arterial are the arterial through volume (two levels), the percentages of turning traffic from the arterial (six levels), the driveway volume (two levels), the percentages of turning traffic from the driveway (three levels), the number of driveways per directions (four levels), the arterial freeflow speed (two levels), and the parking activity frequency (two levels). A more detailed description of the varying factors and levels are given in Table 61 of the corresponding chapter (Chapter 6). It is important to note that since there were two hours of data available for the twolane, twoway simulation model (Park Avenue midday and p.m. peakhour) it was decided to use the p.m. peakhour model for the highlevel arterial through volume and the midday model for the lowlevel arterial through volume. For the twolane, oneway models there were two arterial links, thus the simulation model inputs vary depending on the base model. As such, the first group of scenarios is formed for the Beaver Avenue at Sparks Street, where the simulation input variables are the arterial through volume (three levels), the percentage of turning traffic from the arterial (six levels), the driveway total volume (two levels the values depend on the arterial through volume level), the number of driveways by link (four levels), the arterial freeflow speed (two levels) and the parking activity and duration (two levels). The second group of scenarios corresponds to the base model of Beaver Avenue at Pugh Street and the simulation model inputs are the same as those of the first group plus one additional model input, which is the bus dwell time (two levels). Tables 62 and 63 of the corresponding chapter illustrate the selected values for each level. Database Expansion The next step of the methodology is to perform multiple simulation runs for each different scenario developed. All scenarios created through the factorial design would run for seven times to account for the variability in the simulator. For each new scenario, the appropriate data for the analysis were extracted. First, the midblock delay was calculated based on the following equation: MBD = TT CD LinkLength (3 1) FFS Where: MBD = Midblock delay of each link TT = Arterial link travel time (provided by the simulation output) CD = Intersection control delay at the arterial downstream signal (provided by the simulation output) Link Lengh = The link length of the study arterial link FFS = Arterial freeflow speed (according to the simulation input) Additionally, other information such as the arterial through and turning volume, and the driveway turning volume were extracted from the simulation output. Data Analysis and Formulation of Regression Models The final step of the methodology includes the data analysis and the development of the analytical models. To achieve this, the extracted data and the calculated midblock delay were grouped depending on the arterial configuration under study. Thus, two datasets were created for the case of the twolane twoway arterials and the twolane one way arterials. Each dataset was further divided according to the traffic operations of the arterial to congested and uncongested. The notion behind this decision is that the midblock delay under uncongested conditions is affected significantly by the driveway turning volume and the arterial turning volume which contributes to frequent vehicle frictions; however, under congestion, the midblock delay is primarily a result of the overall congestion. For each of the datasets, one regression equation is formulated, which estimates the calculated midblock delay of Equation 3 1 with respect to the selected independent variables. There were several variables initially considered to be incorporated into the regression models (candidate variables), which are described briefly below. A more detailed description of the candidate variables is included in the relevant chapter of the data analysis (Chapter 6). Vp: Arterial volume (vph/ln): This variable represents the volume that feeds the arterial segment and is measured at the beginning of the arterial link. ZVdr/N: Average Driveway volume (vph/ln): This variable represents the amount of traffic that enters the arterial through the driveways, divided by the number of driveways per link. ZVar/N. Average Arterial Turning Volume (vph/ln): This variable describes the average traffic that exits the arterial through all driveways, either by a leftturn or rightturn maneuver. ZVopp/N: Average Arterial Volume that is Opposed to Arterial Left Turns (vph/ln) (for twolane twoway models): This variable is expressed as the sum of the arterial volume that opposes to left turns from the other direction divided by the number of driveways involved. Xc: Arterial Degree of Saturation at Downstream Intersection: This variable is the arterial demand to capacity ratio measured just upstream of the signal. FFS: Arterial Free Flow Speed (mph) Ndr: Number of driveways per 1000 ft DT. Bus Dwell Time (s) P: Parking Activity (vph/ft): This factor describes the parking frequency of the arterial segment for every available parking space of 20 ft. The final regression models include these variables as they are described above. However, some of them are altered in order to describe more effectively the midblock delay. Last, the midblock delay equations are used for the travel time estimation of two lane twoway or oneway arterials that operate under congested or uncongested conditions according to the following equation: TT = MBD + CD + LinkLength (3 2) FFS Where all measures are as defined earlier. CHAPTER 4 DATA COLLECTION Data collection is an important step of this research and it is used to generate variable arterial traffic conditions in the simulation environment. Thus, sufficient data for the development of the analytical travel time models can be generated. Data Requirements Two types of data are required in this study. The first type includes the input data used for the development of the simulation models: 6. Arterial through volumes. 7. Arterial right and leftturning volumes in the driveways. 8. Driveways turning movements' volumes (right turns, left turns and through movements). 9. % of heavy vehicles on the arterial and on the driveways. 10. Parking and departing maneuvers volumes on the arterials. 11. % time that the segment is occupied by buses when stopped (dwell time). 12. Number of passengers in and out of the buses. 13. Study site geometry (number of driveways per link, driveway spacing, driveway turning radii, number of lanes per link, total link length) 14. Number of bus stops per link and parking bay lengths. 15. Phasing and timing plans for the signalized intersections. The second type of data, namely travel time, were collected concurrently with the input data. The travel time information of the arterials is used for verifying that the developed simulation models replicate efficiently the reallife traffic conditions (model calibration. Description of Study Sites The data collection plan considers two arterials which are located in urban and residential environment. The first one is a twolane twoway arterial (Park Avenue) and the second one is a twolane oneway arterial (Beaver Avenue). The two arterials are shown in the following figures. Both streets are located in State College, PA. Park Avenue does not provide a TWLT lane or left turn pockets, thus each lane per direction serves as a shared lane for through/rightturn and leftturn movements. For Park Avenue two successive links were analyzed. Each link contains two, twoway driveways, which form Tintersections with the arterial. Two separate links are also studied in Beaver Avenue, both of which include a combination of Tintersections and TWSC intersections, with a total of six and eight intersections for each arterial link respectively. Descriptive sketches of the two arterial configurations are presented in Figures 43 through 45. Y JWidw;L~~tlwa Figure 41 Twolane twoway arterial (Park Avenue) Figure 42 Twolane oneway arterial (Beaver Avenue) Data Collection Methods The field data required for this research were collected during peak and nonpeak hours to cover a wide range of flow conditions. 1. Loop detectors and cameras were used for the arterial through vehicle data collection. The loop detectors were located at the approaches of the signalized intersections and also at midblock locations. The cameras were used for those approaches that traffic volumes were not available from loop detectors. 2. Manual recording and cameras were used for the collection of driveway turning and through volumes, the presence of heavy vehicles and the parking activity. 3. Manual recording of the number of passengers that use the buses at the study area and of the respective time the bus decelerates, remains stopped and accelerates at the bus stations. 4. The travel time study was conducted with the floating car technique. The travel time measurements occurred concurrently with the volume measurements. The required number of vehicle runs was calculated based on the standard deviation (S) of the field measured travel time and the margin of error (e) according to the following equation. (1.96 S)2(4 n (4 1) e The calculation was performed for all three study arterials and the summary of the results are given in the following table, where the margin of error was selected at 30 sec. Table 41 Determination of number of vehicle runs based on field measured travel time. PARK B SPARKS PUGH EB WB MID P.M. MID P.M. A.M. MID P.M. A.M. MID P.M. Travel Time (s) 78.6 88.6 112.3 366.0 71.6 75.4 98.9 37.1 35.3 35.4 St.Dev. (s) 14.6 19.1 27.3 55.8 25.2 43.1 18.8 8.9 4.2 5.6 Number of Vehicle Numb7 5 7 4 8 7 7 8 7 7 Runs Performed St.Dev (s) 16.52 39.18 30.51 6.66 Number of Vehicle 1 7 4 1 Runs Required Since the maximum number of required vehicle runs is less or equal to the number of performed vehicle runs, it is concluded that the field measured travel time data are sufficient. Beaver Avenue First Link Sparks Street driveway 1 driveway 2 driveway 3 driveway 4 driveway 5 driveway 6 driveway 7 driveway 8 Figure 43 First twolane oneway arterial link (Beaver Avenue between Sparks St and Atherton St) Beaver Avenue Second Link Pugh Street Pugh Street driveway 1 driveway 2 driveway 3 driveway 4 \/ ~ / ~ '~~\ / driveway 5 driveway 6 Figure 44 Second twolane oneway arterial link (Beaver Avenue between Pugh St and Garner St) Atherton Street Garner Street Park Avenue Successive Links N. Atherton Street driveway 2 N. Allen Rd driveway 3 driveway 4 parking bay driveway 1 Figure 45 Two successive twolane twoway arterial links (Park Avenue between N. Atherton St and N. Allen Rd and Park Avenue between N. Allen Rd and Shortlidge Rd) Shortlidge Rd CHAPTER 5 SIMULATION MODEL DEVELOPMENT An important step of this research is to use simulation to replicate the real time traffic conditions during the data collection study periods, and subsequently use them as the basis for the database expansion. The remaining of this chapter describes the simulation software used, the modeling of the study sites, and their calibration with real field data. Simulation Package For the purposes of this research effort CORSIM (FHWA, 2003) was selected as the appropriate simulation tool for the model development as it can provide control delay in the output. The gap acceptance algorithm of CORSIM and its ability to replicate turning maneuvers are reviewed in the following paragraph. The gap acceptance model in CORSIM is based on default values of the distribution of the acceptable gaps, depending on the driver type category. The program uses 10 behavioral categories and the gap for nearside crossstreet traffic ranges from 56 to 20 tenths of a second, based on a decile distribution. These default values can be altered by the user. The same logic is used for leftturn or rightturn gap acceptance, where the gaps are selected based on the driver characteristics code. If a far side cross street exists, additional gap time is required as input, which depends on the number of lanes. Similarly, the additional time is given by a decile distribution. CORSIM does not have the capability of explicitly defining the speed of a turning maneuver; however, the user can implicitly affect the turning speed of the vehicles on the arterial by defining the freeflow speed of the driveway. Alternatively, it is possible to define the maximum allowable turning speeds in CORSIM, but this option applies to the whole network and not to each individual intersection. The intersection control delay estimation in CORSIM includes the initial deceleration of the vehicles, the stopped delay, and the delay due to acceleration back to the full operating speed. The simulator considers the time difference between the actual travel time of the vehicles versus the travel time had there been no signal (approximated by the ratio of the link length and the operating speed). When the arterial volume is low the operating speed is approximated by the freeflow speed. However, when the flow is significant the operating speed of the vehicle cannot be approximated by the free flow speed and a smaller speed is considered. Although there are other simulation tools available, CORSIM is a broadly used traffic engineering tool and it was selected for this research mostly because it can provide control delay information. Nevertheless, as every software, CORSIM has its own limitations, which will be discussed in a later section. Model Development The simulation of the study sites starts with the input of all the available and appropriate data that were collected in the field. According to the data collection effort there were a total of eight hours of available volume and travel time data; two hours for both twolane twoway arterial links (midday and pm peak hour) and three hours for both twolane oneway arterial links (am peak hour, midday and pm peak hour). Since the two twoway arterial links are adjacent they were modeled into CORSIM together and therefore, each hour was modeled as a separate CORSIM file with the particular volumes. The data used for the simulation model development include geometric data, signal phasing and timing data and volume data. The geometric data for each arterial segment consist of the arterial link length, the location of the driveways, the number of lanes on the arterial and on the driveways, the turning speed into the driveway, the parking space location and the bus turnouts. However, not all geometric data were able to be modeled into CORSIM explicitly due to program limitations. The turning speed of the arterial vehicles on the driveways was modeled by considering the curb radius of the driveways and the driveway throat width. Based on a study performed by Richards (1980) and reported by Stover and Koepke (1988) the speed of a vehicle that enters a driveway is significantly low for all combinations of curb radii and throat widths. The author developed a nomograph that provides the driveway turning speed based on these two measures. Therefore, based on the measure of the curb return radii and the throat width, the turning speed on the driveways was calculated for all cases. However, this type of speed cannot be input directly in CORSIM, as the program allows only the input of a link's freeflow speed. To address this issue, the driveway links were modeled with freeflow speed equal to the corresponding "driveway entry speed" for a length of 100 ft near the intersection. It should be noted that there is an option in CORSIM where the user can specify the maximum allowable turning speed, either left or rightturn, but this option is network wide. Therefore, it is not possible to have different maximum turning speeds on the same arterial street. Additionally, since there is bus activity on one of the arterials, data such as bus frequency and mean dwell time were collected and modeled into the simulator. Although in the study site the buses do not stop at a turnout but at the shoulder lane, this representation could not be done realistically in CORSIM, as the simulator provides only bus turnouts for modeling the bus stops. However, it can be assumed that the effect on traffic operations between a bus stopping at a turnout in CORSIM and a bus stopping at the shoulder lane in the study site is similar. The signal phasing and timing data of the intersections involved were also modeled in CORSIM. All signalized intersections of the study sites operate under semiactuated control. All pertinent data of the semiactuated control such as minimum and maximum green intervals, vehicle passage times, detector locations and size, and offsets were available and modeled appropriately. Moreover, different phasing and timing schemes are available throughout the day for the same intersections, therefore, the modeling of the signal control changed accordingly in all models. At this point it is important to mention that although there were pedestrian phases on the control, pedrelated data were not available. Also, the study sites do not have many pedestrians, but there are a few as the arterials are located in the periphery of the campus. As such, it was assumed that the pedestrian intensity was 20 pedestrians per hour. Additionally, in one case the signal plan considered an exclusive pedestrian phase, but CORSIM cannot model this directly. For this reason, a "dummy" phase replaced the pedestrian phase where the "dummy" vehicle volume would equal the pedestrian intensity and their route would be to a direction that does not interfere with the arterial traffic. Volume data along with proportions of turning movements at the intersections and the driveways and heavy vehicle percentages were also input in the models. Lastly, field measurements of the freeflow speed did not occur in the study corridors thus, it was approximated by the arterial speed limit as in these locations police enforcement of the speed limit is very regular. At this point it should be noted that modeling arterial links into CORSIM requires a large amount of information, only a portion of which were actually collected in the field for this study. For other traffic characteristics such as driver behavior, saturation headway, and startup lost time the default values of the software were initially used, and these were altered as appropriate during the model calibration process. Model Calibration The model calibration process includes first, the comparison of the simulated models with field measurements of travel time, and second, the adjustment of the simulated models that do not match well with the reallife conditions. Once the simulation models were created, the number of the simulation runs needed to be specified to account for the variability in the simulator. As such, several simulation runs were performed and the travel time standard deviation for each arterial model was calculated. By using the following equation and considering a margin of error (e), 15 sec, the number of required simulation runs was calculated as seven. n (1.96*S) (5 1) e The resulting number of simulation runs was very small; thus to simplify the process, the maximum number of required vehicle runs was considered as the number of simulation runs to be performed in every model. Therefore, each of the eight models would run for seven times. The next step of the process is to verify that the simulation models would replicate traffic operations on the arterials and this is accomplished by comparing the average travel times derived from the simulation with the field measured travel times. If the travel time calculated from the simulation would range between 10% of the travel time collected in the field then the respective models would not need any further adjustment. The models that yielded acceptable travel time prediction are the Sparks a.m. and mid model and the Pugh a.m. model. However, if the simulation models yielded travel time outside of the 10% acceptable range they would then be calibrated by making reasonable adjustments in those characteristics that were not collected in the field. Typically, the adjustments that are made in the simulation environment include changes to the saturation (discharge) headway, the mean startup delay, the driver reaction time etc. The models that their predicted travel time was beyond the acceptable range where the Park mid and p.m., the Sparks p.m. and the Pugh mid and p.m. models. The calibration parameters for those models that were adjusted include the mean startup delay, and discharge headway, the time reaction to deceleration, the duration of lane change maneuver, the parking maneuver duration, the driver familiarity, the freeflow speed and the spillback probability of discharging with respect to the discharge position. The default and calibration values of these parameters are given in tables 51 to 53. Table 51 Calibration parameters for Park Avenue midday and p.m. models CORSIM Vl Calibration Parameter C M Calibration Value Default Value 1 (EB), 2.5 (WB) Mean startup delay (s) 2 1 1.5 driveways 1.4 (EB), 2 (WB) Mean discharge headway (s) 1.8 1. 1.5 driveways Probability (%) of a vehicle joining 1 2 3 4 1 2 3 4 spillback with respect to the number of vehicles in the spillback 80 40 0 0 0 5 0 0 Driver familiarity (% drivers that 50 know 1 turn movement in advance) Time reaction to deceleration (s) 1 0.8 Duration of lane change maneuver (s) 3 2 Parking maneuver duration (s) 4 3.5 Table 52 Calibration parameters for Sparks p.m. model CORSIM Calibration Parameter Defat V e Calibration Value Default Value 34 arterial Mean startup delay (s) 2 3 driveways Mean discharge headway (s) 1.8 3 Driver familiarity (% drivers that 80 know 1 turn movement in advance) Time reaction to deceleration (s) 1 3 Duration of lane change maneuver (s) 3 2 % drivers who cooperate with lane 50 20 changes Parking maneuver duration (s) 4 3.5 Table 53 Calibration parameters for Pugh midday and p.m. models CORSIM Calibration Parameter M Calibration Value Default Value Mean startup delay (s) 2 1.5 Mean discharge headway (s) 1.8 1.5 Time reaction to deceleration (s) 1 0.7 Free flow speed (mph) _30 Parking maneuver duration (s) 4 3.5 The probability of a vehicle joining a "spillback" queue with respect to the number of vehicles in the spillback (Table 51) is an important variable which was used for the models that had congestion and yielded high travel times. It was observed that due to the congestion, the arterial through vehicles would not leave gaps, even if the queues were extended beyond the unsignalized intersection. This would lead to an unrealistic representation of field conditions and to increased travel times as the vehicles blocking the intersections would impede the left turns from the opposing direction. In reality, queued vehicles usually leave the unsignalized intersection clear for the driveway vehicles to enter or cross or for the leftturn vehicles on the opposing direction. However, it is possible to define in CORSIM the probability of a vehicle joining a spillback and in this case this probability was reduced (see Table 51) to account for vehicles waiting until the spillback ahead dissipates. The particular combination of probabilities of joining spillback (0% if there is 1 vehicle in the spillback and 5% if there are 2 vehicles in the spillback) resulted in lower travel times for the simulation models, as the vehicles from the opposing directions would leave gaps at the intersections for the leftturning vehicles to cross. By implementing these calibration parameters the travel times from simulation were in agreement with the field measured travel times. The following table presents the simulated and field measured travel times for all models. Table 54 Field measured vs. simulation travel time after calibration Modl Field Measurements Simulation Model . Travel Time Acceptable Range Travel Time Park EB 78.57 [70.71,86.43] 78.16 mid WB 112.29 [101.06,123.52] 104.13 Park EB 88.60 [79.74,97.46] 96.73 pm WB 355.40 [319.86,390.94] 345.76 Sparks a.m. 71.63 [64.47,78.79] 74.76 Sparks mid 75.43 [67.89,82.97] 74.49 Sparks p.m. 98.86 [88.97,108.75] 97.43 Pugh a.m. 37.13 [33.42,40.84] 39.83 Pugh mid 35.29 [31.76,38.82] 36.01 Pugh p.m. 35.43 [31.89,38.97] 33.82 Summary and Conclusions The simulation model development process includes two important steps. The first step was to recreate the geometry of the arterials and the traffic conditions into the simulation environment during the data collection period. The second step was to calibrate the models in order to ascertain that they can replicate satisfactorily traffic operations on the arterials. During the simulation model development several weaknesses of CORSIM were revealed. Although CORSIM is a widely used tool there are some issues that the program does not address directly. For this reason this section also provides some recommendations for improving/extending the capabilities of the software. CORSIM does not provide the option for defining the arterial turning speed for individual intersections. Although the program allows for networkwide input of the maximum right or leftturning speed, it may be more useful that this option is applied to individual intersections based on the research purposes. In this research the issue of modeling right or leftturn maneuvers' speed was addressed by "forcing" the vehicles to enter the driveway with low freeflow speed, which corresponded to the site specifications of throat width and curb radius. Additionally, CORSIM does not allow for modeling exclusive pedestrian phases. In the simulation modeling process this issue was addressed by creating a "dummy" phase for an approach that does not interfere with the arterial network and traffic flow that equals to the pedestrian intensity. Nevertheless, it may not always be possible to create a "dummy" phase for a traffic movement that does not affect the network; thus, this capability of modeling exclusive pedestrian phases should be provided by the software. 40 Another important limitation that was observed is the gap acceptance algorithm that the software applies. As already mentioned, CORSIM builds the gap acceptance algorithm upon the driver characteristics code, but in reality this is not the only basis for the gap acceptance. It is recommended that the software accounts for the fact that drivers become impatient while waiting a long time for a gap. CHAPTER 6 ANALYTICAL MODEL DEVELOPMENT The effects of midblock phenomena are explored with the help of the CORSIM simulation package. The analysis includes two basic steps; the expansion of the available field data set through simulation and the development of the analytical models through regression analysis. For the expansion of the database, alternative scenarios are built in the simulation environment, according to the prespecified design of experiments. These scenarios are formed for both cases of twolane, twoway and twolane, oneway urban arterial streets, thus two subsets of data are generated through CORSIM. These data are eventually used for the midblock delay model development, with the help of MINITAB statistical analysis package. Database Expansion Due to the limited amount of field data available, it was decided to use simulation in order to generate enough data to be used for the analytical model development. After building the models into CORSIM and calibrating them with the field measured travel time data, they were used as a basis for generating more data and expanding the database. For each calibrated model, several simulation model inputs were selected, to be used as varying factors of the factorial design of the experiment. In an effort to attain large variability in traffic conditions and arterial geometry, each factor was designed with different levels. For example, arterial traffic flow ranges between 800 vph/ln and 1000 vph/ln for the twolane, twoway scenarios and between 400 vph/ln and 800 vph/ln for the twolane, oneway scenarios. Each combination of the varying levels of the factors represents a different scenario in the database. The simulation model inputs depend on the calibrated model. Thus, all cases of twolane, twoway arterials derive from the model of Park Avenue (two adjacent links). However, the twolane oneway cases come from both calibrated models of Beaver Avenue at Sparks Street and at Pugh Street; thus, the selected simulation model inputs are different. Additionally, since there are different levels of volume for each scenario, it was decided to match these with the calibrated models that had similar volume throughput (am, mid and pm models). Tables 6 1 and 6 2 summarize the different inputs and their levels that were used for the database expansion. Table 6 1 corresponds to the cases of twolane, twoway arterials and Tables 6 2 and 6 3 correspond to the cases of two lane, oneway arterials. The selection of the simulation model inputs that appear in tables 6.1 through 6.3 is based on (1) the anticipated form of the midblock delay analytical models, and (2) the CORSIM modeling capabilities. That is, the generated data would be used for applying regression analysis and modeling the arterial midblock delay as a function of several parameters of influence. Thus, the simulation model inputs used for generating these data should be such that affect the arterial travel time and midblock delay. Moreover, since the simulation model inputs are in fact CORSIM inputs, the limitations and capabilities of the software should also be considered. As an example, although the actual amount of traffic that exits the arterial through rightturn or leftturn maneuvers is a more straightforward input than the turning percentage at the intersection, this is not feasible to model in CORSIM, since only the turning percentages can be modeled in the simulator. Table 61 Twolane twoway simulation model inputs for database expansion TwoLane TwoWay Arterials Simulation Model Inputs # of Level Value Levels 1 Arterial Through Volume 2 800 1000 VA (vph/ln) % of Arterial Traffic Performing % of Arterial Traffic Performing 20/5 30/5 20/15 20/25 30/15 30/25 2 Right/Left Maneuver VAR /VAL (%) 6 Driveway Total Volume VDR (vph) 200 300 % of Driveway Traffic Performing 3 Right/Left Maneuver at T 3 30/70 50/50 70/30 Intersections VDRR VDRL (%) 4 # of Driveways by Link per Direction 4 0* 1 2 4 5 Arterial Free Flow Speed (mph) 2 30 45 6 Parking Activity Frequency (mph) 2 20 0 Number of Runs per Case 7 Number of Data Points per Link 872 Subtotal 12208 For zero number of driveways the simulation model inputs 2, and 3 do not apply. Table 62 Twolane oneway simulation model inputs for database expansion. First group of scenarios (from Beaver Avenue at Sparks Street) TwoLane OneWay Arterials Simulation Model Inputs # of Level Value Levels 1 Arterial Through Volume 3 600 650 800 VA (vph/ln) Sof Arterial Traffic Performing 20/5 30/5 20/15 20/25 30/15 30/25 2 Right/Left Maneuver VAR/ VAL (%) 6 Driveway Total Volume VDR (vph) 180/ 200/ 220 260/ 300/ 370 * 3 # of Driveways by Link 4 0** 2 3 4 4 Arterial Free Flow Speed (mph) 2 30 45 5 Parking Activity Frequency (mph) 2 20 0 Number of Runs per Case 7 Number of Data Points 220 Subtotal 1540 The driveway total volume depends on the arterial volume and turning percentage; for 600 vph/ln arterial volume and for the first three levels of arterial turning percentages, the driveway volume is 180 vph. ** For zero number of driveways the simulation model input 2 does not apply. Table 63 Twolane oneway simulation model inputs for database expansion. Second group of scenarios (from Beaver Avenue at Pugh Street) TwoLane OneWay Arterials Simulation Model Inputs # of Level Value Levels 1 Arterial Through Volume 3 400 800 1000 VA (vph/ln) % of Arterial Traffic Performing 20 30 20 20 30 30 2 Right/Left Maneuver VAR VAL (%) 6 5 5 15 25 15 25 Driveway Total Volume VDR (vph) 150/ 300/ 300 200/ 400/ 470 * 3 # of Driveways by Link 4 0** 2 3 4 4 Arterial Free Flow Speed (mph) 2 30 45 5 Bus Activity Dwell Time (s) 2 30 0 6 Parking Activity Frequency (mph) 2 20 0 Number of Runs per Case 7 Number of Data Points 448 Subtotal 3136 The driveway total volume depends on the arterial volume and turning percentage; for 400 vph/ln arterial volume and for the first three levels of arterial turning percentages, the driveway volume is 150 vph. ** For zero number of driveways the simulation model input 2 does not apply. Each new scenario created in CORSIM was run seven times to account for the variability in the traffic simulator. Note that the number of runs and the seed numbers of each run is the same as the ones used during the model calibration procedure. Selection of Simulation Output Performance Measures After performing the required simulation runs in CORSIM, it was desired to select appropriate simulation outputs of performance measures, which would be used for the analytical model development in a later step. The selection of the simulation outputs was made with the notion that several of these outputs would be used for the calculation of the midblock delay, while others would represent the regression model's independent variables. As such, the outputs that were extracted directly from CORSIM, for each simulation run are listed in the following table. Table 64 Selected performance measures extracted from CORSIM CORSIM Performance Measures Control delay at downstream signal (sec/veh) Link travel time (sec/veh) Arterial volume by link (vph) Driveway volume (vph) Based on the methodological framework developed, the control delay and arterial link travel time information are used for the determination of the midblock delay (Equation 6 1). MBD = TT CD LinkLength (6 1) FFS Where all measures are as defined earlier. The arterial volume by link and the driveway volume are used as candidate variables for the regression model development. The extracted driveway volume information includes the total volume that enters and exits the arterial link (i.e., the CORSIM output file provides the total outgoing driveway volume and the total incoming arterial volume). Database Organization The expanded database is organized in an appropriate format to be used for the analytical model development. The database is divided primarily into two sets; the two lane twoway dataset and the twolane oneway dataset. This separation is done due to the fact that not all parameters of midblock delay are common to both arterial configurations. Some parameters that influence midblock delay are not the same for both models. The final models estimate the midblock delay that is experienced by arterial through drivers within a single arterial link (between two traffic signals), as a function of the selected independent variables. For the twolane twoway arterials, the midblock delay of Equation (6 1) is calculated for each approach of the arterial link (eastbound/westbound). The generated data (arterial volumes, driveway volumes etc) are transformed into the independent variables to be used for the regression model, and are organized with respect to the approach that they influence the most. A more detailed description of the candidate variables of the model is given in the data analysis section. The total number of datapoints that were generated from the simulation runs and used for the regression models is 4,676 for the twolane oneway arterials and 12,208 for the twolane twoway arterials. Data Analysis The data analysis involves primarily the selection of the appropriate variables that should describe adequately the regression model to be developed, and finally its formation. Additionally, it is essential that the data are organized in a way that all possible conditions are covered; such as congested vs. not congested conditions, and two lane oneway vs. twolane twoway arterials. Selection of Candidate Variables In this step of the data analysis the variables that best explain arterial midblock delay are selected. A detailed description of the candidate variables to be used in the rest of the analysis is presented in the following section. * Vup: Arterial volume (vph/ln) This variable represents the volume that feeds the arterial segment and is measured at the beginning of the arterial link. It is speculated that the arterial upstream volume affects positively the midblock delay; as the amount of traffic entering the arterial increases, the arterial and driveway vehicle interactions are more intense and thus the midblock delay increases. Nevertheless, there is a limit in the influence of the arterial volume on midblock delay. For instance, if the arterial is congested, the amount of traffic that enters the arterial segment is impeded by the downstream queued vehicles, which means that the actual throughput could be less than that under uncongested conditions. Note that when trying to model the arterial midblock delay, the actual arterial throughput is more useful information than the demand, since the latter is not always met (congested conditions). Also, the prevailing throughput can better explain vehicle interactions within a segment than the projected demand obtained from upstream segments. * XVdr/N: Average Driveway volume (vph/ln) This represents the amount of traffic that enters the arterial through the driveways. In the case of twolane oneway arterial this is the sum of all driveways volume divided by the number of driveways per link. In the case of twolane twoway arterials this variable is defined as the sum of traffic that enters the particular direction, divided by the total number of driveways involved. An example of this is illustrated in the following figure. Driveway 1 N   VdrL  VdrR Driveway 2 Figure 61 Sketch of variable lVdrN In Figure 6 1, one could consider that the vehicles traveling eastbound are mostly affected by the leftturning vehicles of driveway 1 and the rightturning vehicles of driveway 2. The through vehicles would likely decelerate to maintain a safe distance from the entering vehicles. With this assumption, the sum of VdrL and VdrR divided by the two driveways would represent the selected variable that affects the midblock delay for the eastbound approach. More generally, the sum of vehicles that enter a specific direction of the arterial from the driveways, divided by the number of driveways involved, would likely influence the midblock delay that the drivers of that direction experience. The following equation illustrates the average driveway volume used for the twolane twoway model. 'Vdr(EB) (VdrL(EB)i + VdrR(EB)i N N 'Vdr(WB) Z(VdrL(WB)i+ VdrR(WB)i) N N Where: Vdr(EB) /N= average driveway volume that affects EB direction. ZVdr(wB)/N = average driveway volume that affects WB direction. VdrL(EB)I = driveway volume that enters the arterial EB direction through leftturn maneuver at the ith intersection. VdrR(EB) = driveway volume that enters the arterial EB direction through rightturn maneuver at the ith intersection. VdrL(WB) = driveway volume that enters the arterial WB direction through leftturn maneuver at the ith intersection. VdrR(WB) = driveway volume that enters the arterial WB direction through rightturn maneuver at the ith intersection. N = Total number of driveways involved within the segment. The effect of this variable is mostly apparent in noncongested conditions, as it can increase the midblock delay that the arterial vehicles experience. In congested conditions, however, this variable may not affect midblock delay significantly, since the arterial vehicles' speed would be most likely very low and they would not decelerate for the oncoming traffic. S Vart/N: Average Arterial Turning Volume (vph/ln) This variable describes the average traffic that exits the arterial through all driveways, either by leftturn or rightturn maneuver. The logic behind this variable is the same as with the average driveway volume as the through vehicles would likely be delayed by the arterial turning vehicles ahead. The average arterial turning volume is the sum of the right turns and left turns from the arterial that are moving to a particular direction divided by the number of driveways that are involved. A schematic illustration of the variable description is shown in Figure 62. Driveway 1 N _VartR VartL _ VartL VartR VartL Driveway 2 Figure 62 Sketch of variable IVar/N In this case, the average arterial turning volume that affects midblock delay in the eastbound direction should be the sum of the leftturning volume at driveway 1 and right turning volume at driveway 2, divided by the two driveways. Similarly, the rightturning vehicles at driveway 1 and the leftturning vehicles at driveway 2 affect the midblock delay that the vehicles of the westbound approach experience. In general, this variable can be described by the following equations: YVart(EB) X(VartL(EB)z + VartR(EB)) N N Vart(WB) (V artL(WB) + VartR(WB)i) N N Where: Vart(EB) /N = average arterial turning volume that affects EB direction. ZVar(WB) /N= average arterial turning volume that affects WB direction. VartL(EB), = arterial leftturning volume that exits the arterial EB direction at the ith intersection. VartR(EB) = arterial rightturning volume that exits the arterial EB direction at the ith intersection. VartL(B), = arterial leftturning volume that exits the arterial WB direction at the ith intersection. VdrR(WB), = arterial rightturning volume that exits the arterial WB direction at the ith intersection. N = Total number of driveways involved within the segment. * ~Vopp/N: Average Arterial Volume that is Opposed to Arterial Left Turns (vph/ln) The main goal of this variable is to capture the influence of the arterial leftturning vehicles on the arterial through vehicles of the opposing direction. This variable is expressed as the sum of the arterial volume that is just upstream of each driveway divided by the number of driveways involved, and this affects the midblock delay of the opposing direction vehicles. Generally, this variable can be expressed as: pp(EB) artup(EB), for WB MidBlock Delay calculation N 2N, pp(B) artp(B)) for EB MidBlock Delay calculation N ENJ Where: XVopp(EB) /N = average EB arterial volume opposed to left turns from WB direction. IVopp(WB) /N = average WB arterial volume opposed to left turns from EB direction. X(Vartup(EB))= sum of the EB arterial volume that is upstream of each ith intersection. X(Vartup(Bj) = sum of the WB arterial volume that is upstream of each jth intersection. Ni = number of driveways that serve EB left turns. Nj = number of driveways that serve WB left turns. * X,: Arterial Degree of Saturation Another variable that is considered for the final analytical model is the degree of saturation of the arterial segment. This variable is estimated based on the HCM methodology (Chapter 16), according to the following equation: X = s(g/C) Where: v = the arterial demand (vph) s = the saturation flow rate (vph) g = the effective green time of the signal (s) C = the cycle length (s) The degree of saturation is a candidate variable for the model because it represents the unmet demand at the traffic signal and thus the arterial congestion. When the degree of saturation increases, it is expected that the midblock delay would increase due to frequent vehicle interactions. * FFS: Arterial Free Flow Speed (mph) The arterial free flow speed is another candidate variable for the analytical model. It is speculated that when the FFS is high, sudden vehicle decelerations due to turning maneuvers would yield additional delays to through vehicles than for lower FFS. This candidate variable may be important for uncongested conditions where vehicles can achieve desired travel speeds, while in congested conditions all vehicles travel with relatively the same (low) speed and thus, the FFS may not be an explanatory variable. * P: Parking Activity (vph/ft) This factor describes the parking frequency of the arterial segment for every available parking space of 20 ft. The parking activity may affect the travel time as the vehicles that perform such a maneuver (either by parking or by leaving the parking space) would cause the arterial oncoming vehicles to decelerate. * Ndr: Number of driveways per 1000 ft An important variable for the regression model is the number of driveways per 1000 ft, which can be considered as the driveway density. It is expected that the more the driveways, the more the opportunities for turning maneuvers, either from the arterial vehicles or from the driveways' vehicles, leading to more chances for vehicle frictions. For the model development, it is assumed that the midblock delay of each direction is affected by the number of driveways that are adjacent to that direction. * DT: Bus Dwell Time (s) The final variable to be considered for the model is the bus dwell time. This variable represents the time that the bus is stopped at the bus stop and it is used only for the twolane oneway model, as this is the only simulation model available with this kind of input. The effect of buses on vehicle delays is considered from the aspect that vehicles generally tend to decelerate when passing through a busstop in the presence of a bus. Apart from the variables described here, several interaction terms were also considered as candidate variables for the regression models. Some of the candidate interaction terms are the interactions between the total arterial leftturning volume (Va.t L) and the total arterial opposing volume (IVopp), (for twoway arterials only),and the interaction between the driveway density (Ndr) with the total arterial turning volume (XVartL+XVartR) or with the total driveway turning volume (XVdrL++VdrR). These interactions were tested in terms of their applicability and their significance. Regression Models For the development of the analytical models, both datasets of twolane oneway and twolane twoway arterials are further divided into congested and noncongested, depending on the conditions under which they operate. The distinction between the two traffic conditions is not an easy task, since the selected criterion should be as clear as possible. In both cases, the criterion for separating the data is the discharge to demand ratio at the downstream arterial signal since this performance measure can effectively distinguish between the two states. A schematic of the measurements required for the calculation of the discharge to demand ratio is illustrated in Figure 63. The measurement of the discharge is the average flow that traverses the downstream arterial segment, VAR (i.e., the segment between the last driveway and the stop line). The demand measurement equals the discharge from the previous segment, VARTH, minus the arterial turning volume, VARTURN, plus the demand entering from the driveway, DDR. ARTERIAL LINK VAR T__ D Downstream Arterial Segment Figure 63 Measurement of discharge to demand ratio. It is expected that in undersaturated conditions, the discharge to demand ratio is near one, since in those cases all the demand is accommodated and the vehicles are discharged unimpeded at the signal. On the contrary, under congestion, the discharge is significantly lower than the demand, and as such the ratio is expected to be significantly lower than one. Thus, all datapoints with a low discharge to demand ratio were considered to represent congested conditions. The threshold for determining the two operating conditions was set by visually inspecting the graphs of the discharge to demand ratio versus the midblock delay that are presented in Figures 64 and 65 (a) and (b). The graph that corresponds to the twolane oneway cases appears in Figure 64 while the twolane twoway graphs for the low volume and high volume data are shown in Figure 65 parts (a) and (b), respectively. As is appears from the graphs that follow, the boundary value was selected to be 0.95, to account for variability of the simulation model. The discharge to demand ratio appears to be slightly higher than 1.00 for some datapoints, which intuitively is incorrect; however this is a result of volume fluctuations during the simulation period. 56 TWOLANE ONEWAY ARTERIALS 050 055 060 065 070 075 080 085 090 095 100 105 110 DischargelDemand Ratio Figure 64 Dataplot of midblock delay vs. discharge to demand ratio for twolane one way arterials. TWOLANE TWOWAY ARTERIALS LOW VOLUME LEVEL 050 055 060 065 070 075 080 085 090 095 100 105 110 DischargelDemand Ratio TWOLANE TWOWAY ARTERIALS HIGH VOLUME LEVEL 160 140 Reduced discharge  upstream congestion Low discharge and 120 demand rates 60 Uncongested * datapoints100 * 050 055 060 065 070 075 080 085 090 095 100 105 110 Discharge to Demand Ratio B Figure 65 Dataplot of midblock delay vs. discharge to demand ratio for twolane two way arterials. (A) Midblock delay and discharge to demand ratio for low volume level. (B) Midblock delay and discharge to demand ratio for high volume level. It is important to note that congestion appeared to take place for different reasons and with different effects. Several datapoints were identified to describe congested conditions although the ratio of discharge over demand would be high, near the threshold of 0.95. This usually occurred when the arterial discharge is low but also the demand measured at the downstream segment is reduced, due to congestion further upstream (i.e., the measurement of the demand is a function of the number of vehicles discharged upstream on the same link; if this value is reduced due to congestion then the demand would also be low). This type of congestion has different effects depending on the type of facility. In both cases of twolane oneway and twolane twoway models the arterial turning volume at the most downstream driveway is replenished by the volume that enters the arterial through the driveway, leading to a discharge rate that approaches demand and for this reason the ratio remains high. However, for the twolane twoway scenarios, another type of congestion is also observed which yields diminishing discharge over demand ratio associated with very high delays. This usually occurred (Figure 66) when the vehicle discharge at the eastbound or westbound upstream segment was impeded by arterial leftturning traffic that was blocking the through movement (due to high opposing traffic). In these cases, the downstream segment of the arterial (where demand and discharge is measured) does not accommodate a large number of vehicles and thus, the discharge to demand ratio is reduced. Furthermore, driveway vehicles cannot enter the arterial eastbound direction as they would also have to perform a leftturn maneuver, but they lose the priority due to the arterial leftturning vehicles or vehicles from the westbound direction. This type of congestion is also referred to as demand starvation. Reduced discharge rate Reduced discharge rate Figure 66 Congestion occurrence in twolane twoway arterials It was also observed that in several cases the discharge to demand ratio was lower than 0.95 but the midblock delay ranged within low levels. These cases typically occurred when the demand at the driveways had very high values and not all of that demand could discharge. However, that does not necessarily mean that the arterial is operating under congested conditions. As such, these datapoints were considered as uncongested conditions. Lastly, it should be noted that a few simulation runs yielded very low discharge to demand ratios, discharge to capacity ratios, long queues on the arterial and extremely large midblock delay. These datapoints represent gridlock and they were eliminated from the database since there is no distinguishable midblock delay in stopandgo traffic. Regression Model for TwoLane OneWay Arterials The distinction of the data in the two datasets yields two different regression models that are presented in the following table, along with the respective ANOVA tables. The derivation of the final independent variables of the models is based on the candidate variables described in an earlier section. The candidate variables were transformed appropriately, in order to achieve a better model fit, depending on their relationship with the midblock delay. Characteristic trends of the independent variables versus midblock delay are plotted in the figures that follow. Table 65 Midblock delay equations for twolane oneway uncongested conditions MidBlock Delay Equation for Uncongested Conditions MBD= 8.08 + 0.266 x eP +1.73 x Nd + 0.261 x FFS + 0.0161 x DT (62) +0.00551 r +0.00488 ~ N N Predictor Coef SE Coef T P Constant 8.0783 0.3026 26.70 0.000 eP*xe 0.265828 0.004175 63.68 0.000 Ndr 1.72951 0.06496 26.63 0.000 FFS 0.261140 0.006249 41.79 0.000 DT 0.016097 0.003633 4.43 0.000 EVd/N 0.005505 0.001351 4.08 0.000 EVa/N 0.004879 0.001010 4.83 0.000 S = 3.104 RSq = 63.8% RSq(adj) = 63.8% ANOVA Table Source DF SS MS F P Regression 6 75569 12595 1307.04 0.000 Residual Error 4441 42794 10 Total 4447 118364 Independent Variables: ep x  exponentiate X. = arterial degree of saturation, 3 = 3.951 Nd  number of driveways per 1000 ft FFS  arterial freeflow speed (mph) DT  bus dwell time (s) EVd/N  average turning volume that enters the arterial link (vph) EVa/N  average turning volume that exits the arterial link (vph) It was decided to use the exponential Xc instead of the candidate variable Xc (arterial degree of saturation) as this variable appeared to describe better to the dataset. The parameter estimate / was obtained with regression through the origin of the independent variable Xc and the dependent variable In(MBD). The result of the regression yielded a parameter estimate / = 3.951. The effect of parking activity was found to be nonsignificant and for this reason it was removed from the final equation of midblock delay (Equation 62). 100 200 300 400 500 Average Arterial Turning Volume Figure 67 Relationship between midblock delay and independent variables for the two lane oneway uncongested model. (A) Average arterial turning volume, XVart/N. (B) Average driveway turning volume, XVdrN UJ 50 * 20 2oI ,,   10 5 1W 150 2U 250 300 Average Diveway Turing Volume 350 4W 450 500U Table 66 Midblock delay equations for twolane oneway congested conditions MidBlock Delay Equation for Congested Conditions MBD=28566.2x "p 0.184xeP" +10.4xNd, +0.359xFFS Demand (63) 172 ,/D 0.262 V, N N Predictor Coef SE Coef T P Constant 285.41 74.50 3.83 0.000 V,, /Demand 66.24 71.40 0.93 0.355 e' 0.1842 0.1646 1.12 0.264 Ndr 10.383 2.097 4.95 0.000 FFS 0.3592 0.1185 3.03 0.003 (EVd/D)/N 172.317 8.314 20.73 0.000 EVaI/N 0.26250 0.02111 12.43 0.000 S = 13.15 RSq = 82.3% RSq(adj) = 81.8% ANOVA Table Source DF SS MS F P Regression 6 178171 29695 171.70 0.000 Residual Error 222 38394 173 Total 228 216565 Independent Variables: V,, /Demand  arterial volume to demand ratio e  exponentiate X, = arterial degree of saturation, 3 = 4.352 Nd  number of driveways per 1000 ft FFS  arterial freeflow speed (mph) (EVd/D)/N  average turning volume that enters the arterial link to demand ratio (vph) EVa/N  average turning volume that exits the arterial link (vph) Similarly to the previous model the exponential Xc was used instead of the candidate variable Xc (arterial degree of saturation). The parameter estimate / was obtained with regression through the origin of the independent variable Xc and the dependent variable In(MBD). In this case, the result of the regression yielded a parameter estimate / = 4.352. In the case of congested conditions the parking activity was found to have no effect on midblock delay; thus it was removed from the respective equation (Equation 63). 63 160 140 o 40 12 0 o 100 150 20 25 OD 350 Average Arterial Turning Volume A 160 140 100 D 1 O230 0 0 5 O 070 0 0 0 1 80distnctono cogestdv  ngeteddt a te lo 60 0 40 20 0 000 010 020 030 040 050 060 070 080 090 100 Average Diveway Volume to Denmand Rtio B Figure 68 Relationship between midblock delay and independent variables for the two lane oneway congested model. (A) Average arterial turning volume, XVat/N. (B) Average driveway turning volume to demand ratio, (XVdr/D)/N. Regression Model for TwoLane TwoWay Arterials The distinction of congested vs. uncongested data at the low volume dataplot (Figure 65 (a)) appears to be straightforward, if one considers the discharge to demand ratio threshold of 0.95. However, the dataplot of Figure 65 (b) appears more complex but this is understandable given the complexity of traffic operations on high volume two way arterials. In Figure 65 (b), the datapoints with a low ratio but also low midblock delay were considered for the uncongested model. The regression analysis performed on both databases of congested and uncongested conditions yields the results presented in the following table. Representative dataplots of midblock delay versus the independent variables of each model are also available. These dataplots are useful for distinguishing trends among the variables, depending on the prevailing traffic conditions. In both congested and uncongested twolane, twoway models the presence of buses is not considered as an independent variable as the simulation input models do not include such a variable. Table 67 Midblock delay equations for twolane twoway uncongested conditions MidBlock Delay Equation for Uncongested Conditions MBD=13.9+0.0126x V +0.126xFFS+0.03100 +0.0128 P" N N (64) Zv V xZV +0.00502 +0.608xNd, +0.105  N 104 Predictor Coef SE Coef T P Constant 13.9070 0.6274 22.17 0.000 Vup 0.0125814 0.0005931 21.21 0.000 FFS 0.125672 0.009947 12.63 0.000 EVa/N 0.030951 0.002189 14.14 0.000 EVopp/N 0.0128054 0.0002931 43.69 0.000 EVd/N 0.005021 0.002875 1.75 0.081 Ndr 0.60799 0.05581 10.89 0.000 (EVrtL* EVo,)/104 0.104880 0.001704 61.56 0.000 S = 6.174 RSq = 64.7% RSq(adj) = 64.7% ANOVA Table Source DF SS MS F p Regression 7 479460 68494 1797.09 0.000 Residual Error 6860 261462 38 Total 6867 740922 Independent Variables: Vp  arterial volume at the beginning of the link FFS  arterial freeflow speed EVa4/N  average turning volume that exits the arterial link (vph) Vopp/N  average arterial traffic that opposes to left turns (vph) EVd/N  average turning volume that enters the arterial link (vph) Nd  number of driveways per 1000 ft (EVrtL* EVopp)/104  Interaction between arterial right turns and opposing volume (vph vph) The parking frequency was found to be not significant in the determination of mid block delay for uncongested conditions in twolane, twoway arterials. 70 60 0 0 50 100 150 200 250 300 350 60 ('VartL* Vopp)/104 40 _ 70 i I o 20  2 0    Figure 69 Relationship between midblock delay and independent variables for the two lane twoway uncongested model. (A) Interaction between total arterial left turning volume and total arterial opposing volume, (EVa L* 2Vopp)/104. (B) Number of driveways per 1000 ft, Ndr. O 1 2 3 4 5 Table 68 Midblock delay equations for twolane twoway congested conditions MidBlock Delay Equation for Congested Conditions V ZV/ ID MBD=17.86.99x " 25.6 d +0.0983xe + +0.250xFFS Demand N (65) +0.00428x ZVr1, +0.0418x ZV,,, 0.0265x ZVd +0.0123x Vp ZV xZV +2.91x N +0.331xP0.0832 104 Predictor Coef SE Coef T P Constant 17.760 1.945 9.13 0.000 Vup/Demand 6.991 2.044 3.42 0.001 Vdr/N*D 25.635 1.291 19.85 0.000 ePxe 0.098332 0.005147 19.11 0.000 FFS 0.24979 0.02408 10.38 0.000 EVatr 0.004277 0.002156 1.98 0.047 1Vati 0.041828 0.002893 14.46 0.000 EVd, 0.026469 0.001888 14.02 0.000 Vopp 0.0123063 0.0004202 29.29 0.000 Nd. 2.9092 0.4439 6.55 0.000 P 0.3307 0.1775 1.86 0.062 (EV,L* EVop)/104 0.083248 0.005995 13.89 0.000 S = 12.94 RSq = 52.9% RSq(adj) = 52.8% ANOVA Table Source DF SS MS F P Regression 7 99395118 90359 540.03 0.000 Residual Error 5283 83291 167 Total 5290 1877241 Independent Variables: Vup/Demand  ratio of arterial volume at the beginning of the link and demand (EVd/D)/N  average turning volume that enters the arterial link to demand ratio (vph) e *x  exponentiate X, = arterial demand to capacity, 3 = 3.977 FFS  arterial freeflow speed EVar  total arterial rightturning volume (vph) EV_1  total arterial leftturning volume (vph) EVd,  total turning volume that enters the arterial link (vph) Vopp  total arterial traffic that opposes to left turns (vph) Ndr  number of driveways per 1000 ft P  parking frequency per 20 ft of available space (vph/ft) (EVatL* EVopp)/104  Interaction term between arterial right turning vehicles and arterial opposing volume (vph2) The arterial degree of saturation was transformed to exponential Xo, and the calculation of the parameter / was done as described previously. The derived parameter estimate in this case is 3.977. 67 The parking frequency for every 20 ft of available parking space is found to be not significant considering a confidence level of 95%; however, as the pvalue slightly exceeds the threshold of 0.05 for that confidence level, it was decided to include the parameter in the final model. 0.2 0.4 0.6 Vup/Demand 0.8 1 160 140 ** 120 100 40 20 0 (EVartL*EVOpp)/104 B Figure 610 Relationship between midblock delay and independent variables for the twolane twoway congested model. (A) Arterial volume to demand ratio, Vup/Demand. (B) Interaction between total arterial leftturning volume and total arterial opposing volume, (Vat_L* XVopp)/104. i~ **~* I~C **+ F~ ~i I 'Q~ T DiscussionDescription of Independent Variables A more detailed description of the selected independent variables and a discussion of their relationship to midblock delay is presented in this section. e ' For the twolane oneway model, the arterial midblock delay under uncongested conditions is found to increase exponentially with the degree of saturation measured at the downstream signal. When the v/c ratio is low, the flow levels even at the downstream signal are low, which means that vehicles do not interact frequently within the arterial segment. But as the v/c ratio increases and queues are formed at the downstream signal, the arterial operation is moving towards more congested conditions. This would mean that the vehicles are traveling in smaller headways and thus, their frequent interactions are expected to increase the midblock delay. However, when a twolane oneway arterial is operating under congested conditions the degree of saturation affects negatively the midblock delay. When already in oversaturated conditions, an increase of the v/c ratio at the downstream signal means that the number of vehicles discharged is increased. This is an indication that the arterial operation moves towards uncongested states, and to reduced midblock delay. In the case of twolane twoway arterials that operate under congested conditions, as the degree of saturation increases then the midblock delay also increases. This is contrary to the congested model of the oneway arterial; however, the two arterial configurations display differences in the trend of the degree of saturation. The arterial degree of saturation is not included in the twolane, twoway uncongested conditions model because it was not found to be significant. V The arterial midblock delay, under uncongested conditions, increases as the feeding volume increases. This means that when the arterial volume is high and it operates under undersaturated conditions, there are more opportunities for vehicle interactions and more maneuvering actions that affect drivers' speed and thus their travel times. On the other hand, under low flow conditions, the midblock delay is reduced. V Demand In congested conditions, the flowtodemand ratio is significantly less than 1. Furthermore, as operations move towards more congested conditions, the actual throughput is reduced, and thus the ratio is also reduced, which leads to an increase of arterial midblock delay as a result of the overall congestion. On the other hand, an increase of the ratio yields more arterial throughput and therefore less congestion and delay for the through vehicles. This variable is used for the congested models of both arterial configurations. * Nd This variable represents the number of driveways involved in the study link per 1000 ft of arterial link length, as it is described in an earlier section. The driveway density is found to be positively related to the midblock delay in all four models; as the number of driveways increases there are frequent occurrences of vehicles that decelerate to either exit or enter the arterial segment; thus the through vehicles may encounter more delay within the link. * FFS The arterial freeflow speed is also included in all final models. Based on the analysis, it was shown that when the arterial freeflow speed increases, then the mid block delay also increases. For example, if the freeflow speed is relatively high, then the through vehicles will have to decelerate more in order to reduce their speed significantly and avoid the turning vehicles that are traveling with low speeds. Y Vdr N The midblock delay depends on the average number of vehicles that enter the arterial segment from the driveways through left or right turns. By increasing the number of incoming vehicles, the delay that the through drivers experience is increased, due to additional vehicle interference. This variable appears in the uncongested models for both cases of twolane oneway and twoway arterials with the same trend. In the twolane, twoway congested model the total number of vehicles that enter the arterial segment from the driveways (XVdr) is considered instead. SVdr /D N The ratio of the driveway turning volume to the respective demand, averaged through the corresponding driveways is found to be an important variable for both congested models. In congested conditions this ratio is reduced as there are not enough available gaps in the traffic stream for the driveway vehicles to discharge. However, an increase of the ratio means that the driveway vehicles have more opportunities to discharge, thus the arterial operation moves towards less congested conditions with overall less midblock delay. Vart N The average number of vehicles that exit the arterial through right or left turns affects positively the midblock delay in uncongested conditions. As the average number of outgoing vehicles increases, there are more decelerations of the following vehicles and thus their delay is also increased. In congested conditions however, where the midblock delay ranges to high levels, the arterial throughput is reduced and consequently the number of vehicles that perform right or leftturn maneuvers reduces. However, as congestion dissipates and the arterial throughput gradually increases, the number of turning vehicles also increases. I IV art L, Z art R In the twolane, twoway arterial model (congested conditions) the total arterial traffic that performs left and right turns is considered instead of the average turning activity. LV * opp N The average arterial volume of the opposite direction that conflicts with leftturning vehicles affects the midblock delay of the through drivers of that direction. This parameter is very important for the twoway model particularly because the arterials under study have only one lane per direction; thus, a vehicle that performs a leftturn maneuver would have to search for available gaps and the following vehicles would have to stop. SVartLx opp 104 The total arterial leftturning traffic is interacted with the total volume from the opposing direction, in the twolane twoway models, for both uncongested and congested conditions. In Equation (64) (twoway uncongested conditions) the interaction impacts positively the midblock delay. Additionally, as the total arterial left turning volume ZVartL, or the total opposing volume Vopp, increase, the midblock delay also increases. In the twoway congested model (Equation (65)) the interaction term parameter has negative value, but the overall effect of the two independent variables involved remains positive. Note that the marginal effect of an arterial turning vehicle remains positive as long as the total opposing traffic is less than 5,000 vph. This assumption is valid if one considers that there is only one lane per direction and a maximum of four driveways involved. Similarly, the total opposing traffic has a positive effect in midblock delay if the sum of the arterial leftturning traffic is less than approximately 1,500 vph. This is reasonable if the boundary condition of four driveways is taken into account. * DT The bus dwell time is included in the analytical model for the twolane oneway only. The analysis shows that the bus dwell time indeed affects the through vehicles' midblock delay, as the arterial through vehicles generally tend to decelerate when passing through a busstop. This variable is not included in the twoway models because buses were not present at the arterials from the field data collection. * P The parking activity influences positively the midblock delay. Increasing parking activity for every 20ft of available parking space yields an increase on the delay that the arterial through drivers experience. This variable appeared to be important only for the congested model of the twolane, twoway arterials. Conclusions In this study four analytical models for estimating arterial midblock delay are presented. These models estimate midblock delay for twolane twoway and twolane oneway arterial streets that operate under both congested and uncongested conditions. Arterial travel time can be estimated as a function of the midblock delay, the intersection control delay and the arterial running time when operating under freeflowing conditions. The final arterial travel time model for all four cases is expressed as: LinkLength TT = MBD + CD + nk(6 6) FFS Where: TT is the arterial travel time (sec/veh). MBD is the midblock delay and it is calculated from equations 62 through 65, according to the arterial configuration and the operating conditions (sec/veh). CD is the intersection control delay which can be obtained from the HCM methodology (Chapter 16) (sec/veh). Link Length/FFS is the arterial running time under freeflowing conditions (ft/(ft/sec)). A descriptive example problem of the calculation of an arterial link travel time is presented in Appendix D. The example problem calculates the midblock delay, the control delay and the arterial running time under freeflowing conditions components of the model for a twolane one way arterial link. Different regression models are used depending on whether the arterial is congested or not. To distinguish between the two conditions the ratio of the arterial discharge over the demand needs to be calculated. This calculation is performed at the arterial segment just upstream of the traffic signal (between the last driveway and the traffic signal). The discharge to demand ratio is expected to be approaching 1 as operations move towards uncongested conditions. This criterion should be used with caution, as it is possible that the discharge to demand ratio is approaching 1, but the measured demand is reduced due to congestion on the upstream segment. In this case the models that correspond to congested conditions should apply. An important conclusion of this research is that both uncongested and congested models of the twolane oneway arterials have better goodnessoffit measures than the twolane twoway models. The equations developed for the twolane oneway models have reasonable R2 values; however in the twolane twoway models the R2 values are lower. This is mostly due to the different degree of complexity between the operations of the two arterial configurations. From the analysis it can be seen that the distinction between the congested and uncongested states is very clear in the case of the oneway arterials and the regression equations fit fairly well to the data. Additionally, the dataplots shown in Figures 67 and 68 are very descriptive of the relationship between the mid block delay and the independent variables. However, the vehicle interactions that take place in the twoway arterials are much more complicated and for this reason the dataplots in Figures 69 and 610 are scattered. Moreover, the twolane twoway models were developed from data taken from two 75 adjacent arterial links; therefore, they also capture the effect of spillbacks, which is a realistic representation frequently observed in the field. CHAPTER 7 CONCLUSIONS AND RECOMMENDATIONS The research conducted for this thesis resulted in some important conclusions and recommendations concerning the modeling limitations of CORSIM and the analytical models of midblock delay. The research findings associated with the modeling limitations of CORSIM are as follows: CORSIM allows for networkwide input of the maximum right or leftturning speed but it does not allow for individual modification of the turning speed at each driveway. The input of the turning speed at each driveway can be an important factor when analyzing the impact of right or leftturning vehicles on the arterial through traffic. In this research this issue was addressed by "forcing" the vehicles to enter the driveway with low freeflow speed, which corresponded to the site specifications of throat width and curb radius. The gap acceptance algorithm of CORSIM is based on the type of sign (stop or yield), the driver characteristic code (measure of aggressiveness) and the number of lanes that the driver has to cross. However, in reallife conditions, these factors are not the only basis for gap acceptance. It is more realistic that the drivers become impatient when waiting for a long time for a gap and this impatience usually reduces their safety margins. Thus, when waiting for a long time for a gap, it is possible that the drivers' aggressiveness increases. The major results of the development of the analytical models that estimate arterial midblock delay are summarized below: In uncongested conditions, the midblock delay is affected by the turning vehicles from the arterial or the driveway. The turning maneuvers increase the vehicle interactions and thus, they force the arterial through vehicles to decelerate, which leads to an increase of their midblock delay. On the other hand, under congested conditions turning vehicles from the arterial do not contribute to the midblock delay, as the arterial speed is already decreased. Also, vehicles turning into the arterial do not affect midblock delay, as they have fewer opportunities to enter the arterial. In these situations, there is high delay on the arterial. In summary, the midblock delay under congested conditions is primarily influenced by the mainline volume and its degree of saturation. Since different variables influence midblock delay under congested and uncongested conditions the two datasets of twolane oneway and twolane twoway arterials were divided in congested and uncongested. By observing the data it was perceived that for the same arterial throughput the arterial was operating under either freeflowing conditions or congestion. The selected criterion for separating the data is the discharge to demand ratio at the downstream arterial signal. When the discharge to demand ratio is higher than 0.95 then the arterial operates under uncongested conditions. When the ratio is below the 0.95 threshold the arterial is congested. However, if the reduced discharge to demand ratio is the result of high demand at the downstream driveway while the overall arterial performance is freeflowing, then these conditions can also be considered as uncongested conditions. It is shown that this criterion conforms well to the data for the twolane one way arterials. For the twolane twoway arterials the distinction is not as clear due to the increased complexity of the system. The final regression models for the two arterial configurations of study are given by the equations below: For twolane oneway arterials under uncongested conditions the midblock delay (MBD) is: MBD = 8.08 +0.266 x ee +1.73 x N +0.261 xFFS +0.0161 xDT +0.00551 dVr +0.00488 "r N N R2 = 63.8% For twolane oneway arterials under congested conditions the midblock delay (MBD) is: V Vd/D _V MBD =285 66.2x "x 0.184xePX +10.4xNd +0.359xFFS 172Vd 0.262 Demand N N R2 = 82.3% For twolane twoway arterials under uncongested conditions the midblock delay (MBD) is: IV IV IV MBD= 13.9 + 0.0126x V + 0.126x FFS + 0.0310 + 0.0128 + 0.00502 d N N N +0.608xNdr +0.105 r,L 104 R2 = 64.7% For twolane twoway arterials under congested conditions the midblock delay (MBD) is: MBD=17.86.99x 25.6 vD + 0.0983xeIPX +0.250 xFFS+ 0.00428x V Demand N V0.1 x V+3 + 0.0418x YV,,, 0.0265x Vd + 0.0123x VPP + 2.91x Ndr +0.331 xP 0.0832 a104 art! opp104 R2 = 52.9% where all variables as described in Chapter 6. An important result of the analysis is that the models for the twolane oneway arterials have reasonable goodnessoffit measure. In the case of the twolane twoway arterials the goodnessoffit measure of the models is low. This is mostly due to the enclosed degree of complexity in twoway operations, which results from the interactions between the two opposing traffic streams. In twoway arterials it is not always possible to isolate traffic operations by movement of direction, especially in the case of twolane arterials, where the left turns have greater influence to the oncoming through vehicles. The final travel time models derive from substituting the midblock delay equations into the equation presented in Chapter 6 (Equation 6 6). Thus, the necessary steps for estimating the arterial travel time with this method are: Calculate the intersection control delay (can be calculated by applying the HCM methodology (Chapter 16)). Estimate the arterial link travel time under freeflowing conditions. Calculate the discharge to demand ratio at the downstream segment of the arterial link under study to determine its operational characteristics define whether it is operating under congested or uncongested conditions and apply the corresponding equation. Refer to Chapter 6 for guidelines regarding the measurement of the discharge to demand ratio. Obtain all pertinent variables that apply to the midblock equation. These may be the turning movements to and from all the driveways, the arterial degree of saturation (measured just upstream of the intersection), the arterial throughput (measured at the beginning of the arterial link) with the corresponding demand, the arterial freeflow speed, the driveway density (per direction of travel), the average arterial volume that opposes left turns from the arterial, the parking frequency for every 20 ft of available parking space, and the bus dwell time. The research findings provide a good insight to the various parameters that affect the travel time within an arterial link, however it is important that the equations presented above are validated with additional field data. Furthermore, it is recommended that the development of travel time models with consideration of midblock delays are extended to other arterial configurations such as fourlane or sixlane twoway arterials. APPENDIX A PHASINGTIMING DIAGRAMS INTERSECTION OF NORTH COLLEGE, PA ATHERTON STREET AND PARK AVENUE, STATE PHASE 2+5 PHASE 2+6 PHASE 4 PHASE 3 Maximum 38 20 Minimum 2 19 10 Yellow 3 3 3 AllRed 2 2 Passage 2 2 Pedestrian 8 7* 10* Memory NonLocking Ped. Recall NonLocking Cycle 1 25 43 7 25 Cycle 2 31 32 7 20 Cycle 3 22 49 7 22 *Upon pedestrian actuation only Plan No. Time Cycle Offset Remarks 1 1:00 Flash 2 6:00 100 sec 86 sec Cycle 1 3 7:00 90 sec 39 sec Cycle 2 4 10:00 100 sec 86 sec Cycle 1 5 16:00 100 sec 68 sec Cycle 3 6 18:00 100 sec 86 sec Cycle 1 Offsets referenced to start of Phase 2+6 yellow _ I C i'~t t INTERSECTION OF PARK AVENUE COLLEGE, PA AND NORTH ALLEN STREET, STATE PHASE 1 PHASE 2 PHASE 3 Maximum 0 0 Minimum 2 0 2 Yellow 3 3 3 AllRed 1 1 Sec/act. 2 Max init. 23 Passage 1.5 5.4 2 To reduce 10 Before red. 23 Min gap 3 Pedestrian 16 16* Memory NonLocking Ped. Recall NonLocking Cycle 1 7 47 20 Cycle 2 7 43 24 Cycle 3 11 43 20 *Upon pedestrian actuation only Plan No. Time Cycle Offset Remarks 1 6:00 Free Cycle 1 2 7:15 85 sec 10 sec Cycle 1 3 8:15 Free Cycle 1 4 11:30 85 sec 10 sec Cycle 2 5 13:30 Free Cycle 1 6 14:15 85 sec 13 sec Cycle 3 7 17:45 Free Cycle 1 8 24:00 Offsets referenced to start of Phase 2 yellow t 4:n r ir ,1 INTERSECTION OF PARK AVENUE AND SHORTLIDGE ROAD, STATE COLLEGE, PA PHASE 1 PHASE 2 PHASE 3 Maximum 0 0 Minimum 2 0 2 Yellow 3 3.5 3.5 AllRed 1.5 1.5 Sec/act. 2 Max init. 16 Passage 1.5 3.6 2 To reduce 10 Before red. 16 Min gap 3 Pedestrian 15 15* Memory NonLocking Ped. Recall NonLocking Cycle 1 7 45 20 Cycle 2 7 45 20 Cycle 3 7 39 26 *Upon pedestrian actuation only Plan No. Time Cycle Offset Remarks 1 6:00 Free Cycle 1 2 7:15 85 sec 22 sec Cycle 1 3 8:15 Free Cycle 1 4 11:30 85 sec 26 sec Cycle 2 5 13:30 Free Cycle 1 6 14:15 85 sec 26 sec Cycle 3 7 17:45 Free Cycle 1 8 24:00 Offsets referenced to start of Phase 2 yellow r,' N f INTERSECTION OF WEST BEAVER AVENUE COLLEGE, PA AND SPARKS STREET, STATE PHASE 1 PHASE 2 I: Max G 0 Min G 3 3 Yellow 3 3 AllRed 1 1 Passage Pedestrian 7 14 Memory NonLocking Ped. Recall Cycle 1 32 25 Cycle 2 60 30 Cycle 3 64 36 Plan No. Time Cycle Offset Remarks 1 1:00 Flash 2 6:00 65 sec Free Cycle 1 3 7:00 90 sec 1 sec Cycle 2 4 9:30 65 sec Free Cycle 1 5 16:00 100 sec 71 sec Cycle 3 6 18:00 65 sec Free Cycle 1 Offsets referenced to start of Phase 1 yellow INTERSECTION OF SOUTH ATHERTON STREET AND AVENUE, STATE COLLEGE, PA WEST BEAVER PHASE 1+6 PHASE 2+6 PHASE 4 Maximum 0 0 Minimum 0 Yellow 3 3 3 AllRed 1 1 Passage 3 Pedestrian 11 19 Memory Max recall Min recall Ped. Recall Cycle 1 20 42 38 Cycle 2 15 35 40 Cycle 3 19 40 41 Plan No. Time Cycle Offset Remarks 1 0:00 100 sec 56 sec Cycle 1 2 7:00 90 sec 32 sec Cycle 2 3 10:00 100 sec 56 sec Cycle 1 4 16:00 100 sec 90 sec Cycle 3 5 18:00 100 sec 56 sec Cycle 1 Offsets referenced to start of Phase 2+6 yellow 11 INTERSECTION OF EAST BEA STATE COLLEGE, PA VER AVENUE AND SOUTH PUGH STREET, PHASE 2 PHASE 4+8 Maximum 22 Minimum 32 Yellow 3 3 AllRed 1.5 2.5 Passage 3 Pedestrian 7 14 Memory Min recall Ped. recall Cycle 1 41 39 Cycle 2 22 23 Cycle 3 58 42 Plan No. Time Cycle Offset Remarks 1 3:00 Flash 2 6:00 80 sec 3 sec Cycle 1 3 7:00 45 sec 20 sec Cycle 2 4 10:00 80 sec 3 sec Cycle 1 5 16:00 100 sec 75 sec Cycle 3 6 18:00 80 sec 3 sec Cycle 1 Offsets referenced to start of Phase 2 yellow W: t t I INTERSECTION OF EAST BEA STATE COLLEGE, PA C VER AVENUE AND SOUTH GARNER STREET, PHASE 2 PHASE 4+8 Maximum 18 Minimum 37 3 Yellow 3.5 3.4 AllRed 1.5 1.6 Passage 3 Pedestrian 15 14 Memory Min recall Ped. recall Cycle 1 52 28 Cycle 2 59 31 Cycle 3 63 37 Plan No. Time Cycle Offset Remarks 1 0:00 80 sec 33 sec Cycle 1 2 7:00 90 sec 56 sec Cycle 2 3 10:00 80 sec 33 sec Cycle 1 4 16:00 100 sec 90 sec Cycle 3 5 18:00 80 sec 33 sec Cycle 1 Offsets referenced to start of Phase 2 yellow W: _LL _TT APPENDIX B TURNING MOVEMENT AND LOOP DETECTOR DATA PARK AVENUE, 04/20/2004, 11:45AM12:45PM INTERSECTION DATA ST: Atherton Stre t ST: Park Avenue ST: Atherton Stre t ST: Prk AenueST: ark venu ST: Park Avenue SEastbound Westbound Time Left Thru Right Truck Total Left Thru Right Truck Total Total Int. 11:4512:00 0 0 0 0 0 33 1 92 9 126 659 12:0012:15 0 0 0 0 0 45 0 85 7 130 715 12:1512:30 0 0 0 0 0 47 1 136 8 185 750 12:3012:45 0 0 0 0 0 32 1 94 4 127 664 SUM 0 0 0 0 0 157 3 407 28 568 % 0.28 0.01 0.72 0.05 ST: Alien Street ST: Allen Street Northbound Southbound Time Left Thru Right Truck Total Left Thru Right Truck Total 11:45 12:00 13 3 18 1 34 0 1 4 1 5 12:00 12:15 26 2 30 2 57 1 1 1 0 3 12:15 12:30 25 2 20 1 47 0 1 0 0 1 12:30 12:45 11 0 19 3 31 1 1 0 0 2 SUM 75 7 88 7 169 2 4 5 1 11 % 0.44 0.04 0.52 0.04 0.19 0.38 0.43 0.09 ST: Park Avenue ST: Park Avenue __ Eastbound Westbound Time Left Thru Right Truck Total Left Thru Right Truck Total Total Int. 11:45 12:00 2 95 4 3 101 33 112 0 5 145.01 284 12:0012:15 0 102 11 4 113 25 107 1 4 132 306 12:15 12:30 6 73 11 2 90 20 156 0 12 176 314 12:30 12:45 2 104 10 3 116 24 118 1 6 143 292 % 0.02 0.89 0.09 0.03 0.17 0.83 0.00 0.05 Time Left Thru Right Truck Total Left Thru Right Truck Total 11:45 12:00 0 212 35 9 247 88 198 0 7 286 12:00 12:15 0 249 34 8 283 97 204 1 9 302 12:1512:30 0 223 32 16 255 99 211 0 8 310 12:30 12:45 0 199 42 15 241 99 196 0 6 295 SUM 0 883 144 47 1027 384 809 1 29 1194 % 0.00 0.86 0.14 0.05 0.32 0.68 0.00 0.02 10 374 36 I 13 I 420 102 493 2 I 28 I 597 ST: Atherton Street ST: Atherton Street Northbound Southbound SUM ST: Shortlidge Road Northbound ST: Grove Alley Southbound Time Left Thru Right Truck Total Left Thru Right Truck Total 11:45 12:00 15 0 10 8 25 1 0 0 0 1 12:00 12:15 14 0 8 5 22 1 0 1 0 2 12:15 12:30 16 0 10 5 27 0 0 0 0 0 12:30 12:45 11 0 11 4 21 0 0 1 0 1 SUM 57 0 38 22 95 2 0 2 0 4 % 0.60 0.00 0.40 0.23 0.46 0.00 0.54 0.00 ST: Park Avenue ST: Park Avenue Time Left Thru Right Truck Total Left Thru Right Truck Total Total Int. 11:45 12:00 1 100 13 4 113 13 133 0 7 146 285 12:00 12:15 0 125 6 5 131 5 121 0 4 126 281 12:15 12:30 0 82 10 3 93 13 164 0 14 177 296 12:30 12:45 0 117 8 3 126 13 135 1 8 150 298 SUM 1 424 38 15 463 44 553 1 33 598 % 0.00 0.92 0.08 0.03 0.07 0.92 0.00 0.06 DRIVEWAY DATA FIRST LINK: Between Atherton and Alien Driveway 3 (Lischer Rd) Time EBR WBL NBL NBR veh IHV veh HV veh HV veh HV 11:4512:00 30 4 0 8 2 12:0012:15 23 1 2 2 12:1512:30 15 2 0 6 2 12:3012:45 30 3 1 1 7 TOTAL 98 11 3 27 Driveway 4 (N. Borrowes Str) Time EBL WBR SBL SBR veh HV veh HV veh HV veh HV 11:4512:00 2 2 2 1 3 12:0012:15 6 0 0 1 12:1512:30 2 1 1 2 1 2 1 12:3012:45 4 2 0 1 TOTAL 14 6 6 8 SECOND LINK: Between Alien and Sortlidge Driveway 1 (McKee) Time EBL WBR SBL SBR veh HV veh HV veh HV veh HV 11:4512:00 1 5 1 5 12:0012:15 3 1 4 3 1 4 12:1512:30 5 5 1 5 5 1 12:3012:45 4 4 4 4 TOTAL 14 19 11 8 I ST Park Avenue ST: Park Avenue Eastbound Westbound Driveway 2 (Hohles) Time EBL WBR SBL SBR veh HV veh HV veh HV veh HV 11:4512:00 1 1 0 0 12:0012:15 1 3 1 2 12:1512:30 1 5 2 1 1 12:3012:45 0 2 1 0 TOTAL 3 11 4 4 LOOP DETECTOR DATA BURROWES & ALLEN EB WB 11:4512:00 101 129 12:0012:15 113 128 12:1512:30 90 186 12:3012:45 116 130 TOTAL 420 573 LISCHER & ATHERTON EB WB 123 126 140 130 99 185 143 127 505 568 HOLMES & SHORTLIDGE EB WB 113 148 131 134 93 184 126 148 463 615 TURNING MOVEMENT AND LOOP DETECTOR DATA FOR PARK AVENUE, 04/20/2004, 04:30PM05:30PM INTERSECTION DATA ST: Atherton Stre t ST: Atherton Street' ST: AthertonStree ST: Atherton Stre t Time Left Thru Right Truck Total Left Thru Right Truck Total 04:3004:45 0 314 45 3 360 94 256 0 8 350 04:4505:00 0 326 61 4 386 78 243 0 6 321 05:0005:15 0 343 70 9 413 69 240 0 5 309 05:1505:30 0 376 65 6 440 73 243 0 2 316 SUM 0 1359 241 22 1600 314 983 0 21 1297 % 0.00 0.85 0.15 0.01 0.24 0.76 0.00 0.02 Time Left Thru Right Truck Total Left Thru Right Truck Total Total Int. 04:3004:45 0 0 0 0 0 37 2 148 5 187 897 04:4505:00 0 0 0 0 0 47 4 164 8 214 922 05:0005:15 0 0 0 0 0 41 1 160 7 202 924 05:1505:30 0 0 0 0 0 36 1 159 4 196 953 SUM 0 0 0 0 0 160 8 631 24 799 % 0.20 0.01 0.79 0.03 : ar venu : ar venue e I Northbound Southbound cT. P klr Airn..fl CT. P klr Airn..a Eastbound Westbound 