Click for next page ( 40


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 39
APPENDIX A LITERATURE REVIEW Traff~c-actuated control has been used since the early i930's. The operational performance of isolated actuated controlled intersections or a set of coorclinated actuated intersections largely depends on traffic patterns and actuated controller parameters. A wed-designee! actuated control plan that responds appropriately to traffic demand can significantly reduce delay and fuel consumption. Therefore, shortly after actuated signal control was first introduced, researchers began to study the influence of traffic arrival and departure characteristics at a signalized intersection with traffic- actuated control. Many researchers also focused on optimizing the controller settings, detector placement, and the relationship between them. Although more advanced forms of adaptive traffic control strategies were introduced recently, the original concepts of traffic-actuated control still play a very important role today. Review of past and current research is an area that definitely merits attention. There has been a substantial amount of research conducted on traff~c-actuated control that will contribute to this project. The main topics of this literature review include: . Traffic-actuated control definitions, Warrants for traffic-actuated control, Benefits and operating considerations for traffic-actuated control, Effects of coordination and phase-skipping for traffic-actuated control, Late-night, low-volume operation of coordinated actuated systems, Evaluation of traffic-actuated control by simulation, Estimation of green times and cycle length for traffic-actuated control, Delay models for traffic-actuated control, Signalized intersection capacity models for traffic-actuated control, and Overview and evaluation of "Enhancement of the Value Iteration Program for Actuated Signals" (EVIPAS). The procedure contained in the Highway Capacity Manual (HCM) [~] is used almost as a standard to analyze signalized intersection capacity and level of service, so the major emphasis of this review is the application to capacity and level of service analysis. One item that needs to be reviewed carefully is a new program called EWPAS. EVIPAS is an optimization program that analyzes and determines the optimal settings of timing parameters for traffic-actuated control. The results of the testing efforts on EV1PAS are reported at the end of this chapter. TR\FFIC-ACTUATED CONTROL DEFINITIONS Three basic forms of traffic control: pretested, actuated and semi-actuated were mentioned by Orcutt [2] in 1975. He indicated that pretimed control was used primarily in the Central Business District AppendixA: Page 1

OCR for page 39
(CBD) area, especially where a network of signals must be coordinated. He defined actuated signals in terms of equipment that responds to actual traffic demand of one or more movements as registered by detectors. If all movements are detected, the operation is referred to as "fully-actuated." If detectors are installed for some, but not all, traffic movements, the term "semi-actuated" was applied. Orcutt suggested fi~ll-actuated control should normally be used at isolated intersections. Precise definitions of the basic controller types were described by the National Electrical Manufacturers Association ENEMA) standards [3] in 1976. According to the NEMA standards, the basic controllers include pretimed, semi-actuated, full-actuated without volume-density features, and full-actuated with volume-density features. In the remainder ofthis discussion, the three types of traffic-actuated control will be designated as sem~-actuated' fill-actuated and volume-density control. WARRANTS FOR TRAFFIC-ACTUATED CONTROL Warrants for selecting traffic control modes, which are very useful for practicing engineers, have been researched since the early 1960's. Studies of delay at actuated signals have been made with the pur- pose of evaluating warrants for this type of control based on the 1961 Manual on Uniform Traffic Control Devices (MUTCD) information. This information was expanded by the Texas Department of Highways and Public Transportation into a graphical format. The graphical relationships were studied in 1971 by Vodrazka, Lee and Haenel [4], concluding that they provide good guidelines for selecting actuated equipment for locations where traffic volumes do not warrant pretimed signals. The current edition oftheMUTCD stops short of numericalwarrants for choosing between pretimed and traff~c-actuated control, but it does suggest certain qualitative conditions under which traffic- actuated control should be implemented. In 1967, Gerlough and Wagner [5] began to compare pretimed control effectiveness with volume- density control. They found that traffic-actuated control at higher traffic volumes degraded performance. One of the problems cited for volume-density control was that the duration of green for each phase was dependent on the estimated queue length at the beginning of the phase. Difficulties with queue length estimation made this type of control less effective. BENEFITS AND OPERATING CONSIDERATIONS FOR TRAFFIC-ACTUATED CONTROL Long-loop presence detection operates by producing a vehicle call for the duration of time that the vehicle is over the detector. On the other hand, small-area detection mode operates by producing a pulse of less than 0.1 seconds when the vehicle is first detected. The latter mode of operation is known as passage' pulse or count detection. The long-loop presence detector with full-actuated controllers is known as lane-occupancy control or loop-occupancy control BLOC). LOC operation occurs when the controller is programmed for initial green interval of zero. Extensions are set either to zero or to a very low value. There is no need for a non-zero initial interval or minimum green time because the long loops continuously register the presence of vehicles that are waiting' causing the AppendixA: Page 2

OCR for page 39
controller to extend the green until the entire queue is discharged. The result is a signal operation that responds rapidly to changes in traffic flow. In 1970, Bang and Nilsson t6] compared LOC operation with small area detector (pulse detector) operation. They concluded that delay was reduced 10 percent ant! stops by 6 percent under the same traffic conditions with LOC. In 1975, Cnbbins and Meyer t7] compared pulse and presence detectors. They concluded that the longer the length of the presence detector on the major approach to the intersection, the longer the delay. They also concluded that the highest intersection travel time values occurred when either long-Ioop presence or pulse detectors were used on both major and minor approaches. The Intersection travel time was cleaned here as the average time it takes a vehicle to pass through an intersection, whether it is stopped or slowed. Numerous theoretical studies on traffic signal timing were conducted between 1958 and 1970. The theoretical work on pretimed control by Webster [~] in 1958 and Miller [9] in 1963 has been applied to the computation of optimum cycle lengths as a function of vehicle amval rates. It has also been used for evaluating vehicle delay, intersection capacity, probability of stops and so on. These results were also wed validated through the comparison of field data. In 1969, Newell tI0] and Newell and Osuma [l I] expanded the body of theory by developing relationships for mean vehicle delay with both Retimed and actuated control at intersections of one-way streets and intersections oftwo-way streets, respectively. In reference 10, Newell demonstrated that the average delay per vehicle for an actuated signal is less than that of a pretimed signal by a factor of about three for intersections of one- way streets. Reference Il considered intersections oftwo-way streets without turning vehicles. For the particular vehicle-actuated policy which holds the green until the queue has been discharged, the vehicle-actuated control will not perform as well as pretimed control when I) flows are nearly equal on both approaches of a given phase, and 2) the intersection is nearly saturated. This work confirms Bang's results, previously described. In 1976, Staunton [12] summarized the work of numerous signal control researchers. In his paper, the comparisons of delay produced by pretimed and actuated control, as a function of vehicle volumes, were presented. Staunton demonstrated that fi~-actuated control with 2.5 see extensions will always be better than the best form of pretimed operation, given optimum settings for all volumes. Longer values for the extensions can easily degrade actuated control performance. His conclusions were based on simulation, but the details ofthe detector configuration were not specified. In view of the 2.5 see extension time, passage detectors were probably used in this study. The performance estimates from Staunton were supported by Bang [133. In 1981, Tarnoff and Parsonson [14] compiled an extensive literature review on the selection of the most appropriate form of traffic control for an individual intersection. Three complementary approaches were used to evaluate controller effectiveness: I) field data collection using observers to manually measure vehicle volumes, stops and delay; 2) simulation using the NETSIM mode} developed by the Federal Highway Administration (FHWA) to evaluate control system performance; and 3) analytical techniques developed by the research team and other agencies. Following is a list of the more relevant conclusions from their extensive literature review. AppendixA: Page 3

OCR for page 39
I. Pretested controllers operate most effectively when the shortest possible cycle length is used subject to the constraints of providing adequate intersection capacity and minimum green times for pedestrians and vehicle clearance intervals. 2. 4. The delay produced by fit actuated controllers is extremely sensitive to the extension time setting used. In general, shorter gap extensions reduce vehicle delay. For small area detectors ~ motion or pulse detectors), at low and moderate volumes when extensions oftwo or three seconds are employed, the use of fit actuated con- trollers will produce reduced delays and stops over those obtained using pretimed controllers. When high traffic volumes occur on both main and side streets causing the controller to extend the green time to the maximum on all phases, the fi~li-actuated controller wait perform as a pretimed controller, producing comparable measures of vehicle flow. The relative effectiveness of the various control alternatives depends on the quality of the signal timing employed. A poorly timed actuated controller will degrade traffic performance to as great an extent as a poorly timed pretimed controller. Through the detailed evaluation of controller performance, the conclusions by Tarnoff and Parsonson [14] are descnbed as follows: Sern~-actuated controllers produce a higher level of stops and delays for ad traffic conditions than either the full-actuated or pret~med controllers. However, for side streets operating traffic volumes less than 20 percent lower than main street volumes, the difference in effectiveness between sem~-actuated and fi~-actuated controller is insignificant. FuD-actuated controllers operating in an eight-phase dual-ring configuration produce significant operating benefits over the four-phase pretimed controller operation. Simulation results on small area detectors showed that for approach speeds of 3 5 mph and a 2-second vehicle interval, locating the detectors farther apart Dom the stop line (150 h) produced a much better performance level. Tarnoff and Parsonson concluded that it is appropriate to locate the detector such that the travel time is equal to the extension time. It was also concluded that for volumes in excess of 450 vehicles per hour per lane and approach speeds of 35 mph or higher, additional improvement can be achieved through the use of the added initial volume-density controller feature. They indicated that further moclest gains in performance for full-actuated control were possible with the use of long loops and short (or zero) initial and extension settings. This application was found to produce a performance similar to a two second extension time with a short loop. From the s~rnulation results, they also concluded that LOC was more effective than pulse detection over a wide range of traffic volumes. LOC offers another advantage not reflected by the simulation results: _ ~ . . . . . . . .. . , . . ~ screening out false calls caused by vehicles approaching but not traveling through the intersection. Volume-density controllers provide the greatest benefit at intersections with high approach speeds and detector setbacks in excess of 125 ~ from the intersection. This situation requires a variable initial green time. Tarnoff and Parsonson found that the t~me-wait~ng and extension-reduction options AppendixA: Page 4

OCR for page 39
ofthe volume-density control clid not improve the controllers perfonnartce over that of a fi~-actuated controller unless the option is used to reduce the vehicle extension to a value that is less than the one used for the fi~-actuated controller. Thus, if the volume-density controller is timed to provide a three-see passage time and a two-see minimum extension, its performance wait be superior to that of a fi~-actuated controller with a constant three-see extension. It should be noted that Tarnoff and Parsonson's simulation results do not properly account for the problem of premature termination of green resulting from the variation in oueue discharge headwavs that occur under normal operating conditions. In 1985, Lin [15] studied the optimal timing settings and detector lengths for full-actuated signals operating in presence mode using the RAPID simulation model. He suggested that the optimal maximum green for hourly flow patterns with a peak hour factor of I.0 was about ten see longer than the corresponding optimal greens with a peak hour factor of 0.85, and that the optimal maximum green was approximately 80 percent longer than the corresponding optimal greens. This result was similar to the I.5 times pretimed split suggested by Kell and Fulderton [161. I~in indicated that optimal vehicle intervals were a function of detector length and flow rate. For detectors 30, 50 and 80 Bc long, the use oftwo-second, one-second and zero-second vehicle intervals can lead to the best signal performance over a wide range of operating conditions, respectively. The use of vehicle intervals greater than zero see for detectors 80 feet or longer is not desirable unless the combined critical flow at an intersection exceeds 1,400 vph. In order to improve the V]PAS model, a new optimization algorithm and a new intersection simulation were designed and programmed. The original V1PAS traffic characteristics and vehicle generation routines were combined with these new models to create the enhanced version called EVIPAS. In 1987, Bullen, Hummon, Bayer and Nekmat [17] developed EVIPAS, a computer mode} for the optimal design of a vehicle-actuated traffic signal. - The EVIPAS mode] was designed to analyze and optimize a wide range of intersection, phasing, and controller characteristics of an isolated, fi~ly-actuated traffic signal. It can evaluate almost any phasing combination available in a two to eight-phase NEMA type controller and similar phasing structures for a Type 170 controller. They indicated that the mode} can provide optimum timing settings for pretimed, sem~-actuated, fi~- actuated, or volume-density control using a variety of measures of effectiveness chosen by the users. The mode} has been field tested and validated. In 1987, Messer and Chang tIS] conducted field studies to evaluate four types of basic full-actuated signal control systems operating at three diamond interchanges. Two signal phasing strategies were tested: a) three-phase and b) four-phase with two overlap. Two small-Ioop (point) detection patterns (single- and multi-point) were evaluated for each type of phasing. They concluded that I) single- point detection was the most cost-e~ective three-phase design; and 2) multi-point detection was the more delay-effective four-phase configuration. Four-phase control characteristically operates a longer cycle length than the three-phase for a given traffic volume. This feature may produce higher average delays unless the cycle increase is controlled to the extent that the internal progression features of four-phase can overcome this deficiency. AppendlixA: Page 5

OCR for page 39
In 1989, Courage and Luh [19] developed guidelines for determining the traffic-actuated signal control parameters which would produce the optimal operation identified by S0AP84. They also evaluated the existing signal control parameters on an individual traffic-actuated signal. The significant conclusions are summarized as follows: Under low volumes, the maximum green settings have little or no effect on the performance of actuated signal controllers. Under moderate volumes, shorter maximum greens increase the average delay considerably. Longer maximum greens, however, have no significant effect on delay. Under high volumes, the maximum green settings become more important. There is a setting which minimizes average delay. Other settings with longer or shorter greens will produce more average delays. The optimal maximum green setting can be achieved by running SOAP under actuated control with an optimal saturation level set in the BEGIN card. They indicated that the settings which are optimal at some time may not be appropriate in other times of day. The value of 0.95, which is the default value used in SOAP actuated control, was suggested for multiphase operation and a slightly higher saturation level may be desirable for two phase operation. For approaches with a reasonably even distribution of traffic volume by lane, settings of 4.0, 2.0 and 1.4 seconds were recommended for 1, 2 and 3 lane approaches as the best values for extension, respectively. In the same year, Bullen [20] used the EV]PAS simulation and optimization model to analyze traffic- actuated traffic signals. The variables studied were detector type and placement and the settings were minimum green and vehicle extension. The evaluation criterion was minimum average vehicle delay. The study showed that the optimum design of a traffic-actuated signal was specific for some variables but relatively unaffected by others. The design was critical only for high traffic volumes. At low volumes, vehicle delay is relatively unaffected by the design parameters studied in his paper. The most critical variable Bullen found was vehicle extension, particularly for passage detectors, where it should be at least 4.0 seconds regardless of detector placement and approach speed. This conclusion somewhat contradicts previous study results. However, it should be noted that the model (EV1PAS) used by Bullen considered variable queue discharge headways. Detector configuration is essential to the success of actuated control. Kell and Fullerton [21] in their second edition ofthe Manual of Traffic Signal Design in 1991 indicated that the small area detector might ideally be located three or four seconds of travel time back from the intersection, with the allowable gap set accordingly. Similar principles were proposed in previous research by Tarnoff and Parsonson. Kell and Fullerton also indicated that, in some states, the detector setbacks were determined based on the safe stopping distance. The main purpose is to avoid the dilemma zone in which a vehicle can neither pass through the intersection nor stop before the stop line. For long loop detectors, they indicated that the concept of 100D occupancy control can provide good operation ~ ~. 1' ~1 ~' . , ~. _ when vehicle platoons are wen torrnea. 1ne use of several smaller loops instead of one long loop was ~ . ~ .1 1 ~ ~ ~ suggested to solve the problem or random vehicles causing excessive green. In 1993, Bonneson and McCoy [22] proposed a methodology for evaluating traffic detector designs. They indicated that the safety and efficiency of a traffic detector design can be determined by the Appendix A: Page 6

OCR for page 39
probability of max-out and the amount of time spent waiting for gap-out and the subsequent phase change. The stop line detector and advance loop detector with presence and pulse mode were discussed, respectively. The methodology presented by Bonneson and McCoy determined the optimal combination of design elements in terms of safety (via infrequent max-out) and operations (via a short waiting time for phase change). The design elements included detector location, detector length, vehicle speed, passage time settings, and call extension setting. They concluded that a large maximum allowable headway will have an adverse effect on performance by increasing the max-out probability and the length of wait for phase change. EFFECTS OF COORDINATION AND PHASE-SKIPPING FOR TRAFfIC-ACTUATED CONTROL . In 1986, lovanis and Gregor [23] studied the coordination of actuated arterial traffic signal systems. In the past, all existed optimization methods required that each actuated signal be converted to its nearest equivalent pretimed unit. Using bandwidth max~rruzation as a starting point, a new procedure was developed by Jovanis and Gregor that specifically accounts for actuated timing flexibility. Yield points and force offs at non-cntical signals are adjusted so they just touch the edges of the through- band while critical signals are unmodified. This method was applied to a data set describing m~d-day traffic conditions on an urban arterial system of six signals in central Illinois. Simulation was used to evaluate these timing plans and compare them with corresponding pretimed alternatives. They were surprised to find out that pretimed, coordinated control appeared superior in general to actuated coordinated control in this experiment. They also concluded that the level of service ~OS) of side streets were much more important for pretimed than actuated strategies. In 1989, Courage and Wallace [24] developed guidelines for implementing computerized timing designs from computer programs such as PASSER Il. TRANSYT-7F and AAP in arterial traffic control systems. The coordination of a group of traffic-actuated signals must be provided by some forest of supervision which is synchronized to a background cycle length with splits and offsets superimposed. Both external and internal coordination of the local controllers were addressed. This report focused on the external coordination oftraffic signal controllers. Permissive periods were introduced to indicate the time interval following the yield point during which the controller is allowed to yield to cross street demand. If the computed splits are longer than the minimum phase times, it might be possible to establish a permissive period without further sacrifice or compromise on the rest ofthe sequence. The methodology of computing permissive periods was described. The effect of phase-skipping due to lack of traffic demand was also presented. The Timing Implementation Method for Actuated Coordinated Systems (TIMACS) program was developed to perform permissive period computations. Since most previous studies were more specific to certain geometric and phasing combinations, the qualitative and quantitative evaluation methodology for coordinated actuated control needed to be fully investigated. In 1994, Chang and Koothrappally t25] designed a field study to demonstrate the operational effectiveness of using coordinated, actuated control. They concluded: I) There was AppendixA: Page 7

OCR for page 39
significant improvement, based on both delay and number of stops, between the semi-actuated control, full-actuated, and pretimed coordinated timing during the study; 2) There were no significant differences In performance among all the sem~-actuated operations as long as the progression-based signal coordination timing was developed correctly; and 3) The use of longer coordination cycle lengths generally caused less arsenal stops. However, it would generate much higher overall system delays. LATE-NIG~:T~ LOW-VOL~ OPERATION OF COORDINATED ACTUATED SYSTEMS Coordinating the timing of adjacent signals to promote progressive traffic movement was recognized as one ofthe most effective means for reducing vehicular stops, delay, fuel consumption, and exhaust emissions. Early efforts on the subject of signal control always indicated the need to interconnect signals into a single system and to work toward maximizing progressive movement during peak periods. In 1990' Luh and Courage [26] evaluated the late-night traffic signal control strategies for arterial systems. They stated that late-night, low-volume arterial signal control involved a trade-off between the motorists on the artery and those on the cross street. The conventional measures of effectiveness such as stops, delay, and fuel consumption were not appropriate for evaluating this trade-off. Luh and Courage proposed a methodology to choose between coordination and Dee operation on arterial roadways controlled by semi-actuated signals when traffic is light. The choice was made on the basis of a disutility Unction that was a combination of the number of stops on the artery and the average cross-street waiting time. The results indicated that this method provides a promising tool for late-n~ght arterial signal control. EVALUATION OF TRAF1?IC-ACTUATED CONTROL BY SIMULATION Simulation modeling has become an extremely important approach to analyzing complex systems. After 1980, more and more simulation modeling was used in traffic operations. In 1984, Lin and Percy [27] investigated the interactions between queuing vehicles and detectors for actuated controls, which govern the initiation, extension, and termination of a green duration. They emphasized that the model used in the simulation analysis should be calibrated in terms of observed characteristics such as queue discharge headway, arrival headway, the relationship between the arrival time of a queuing vehicle and the departure time of its leading vehicle, the number of queuing vehicles in a defined area at the onset of a green duration, and the dwell time of a vehicle on the detection area. They also indicated that under a presence control, the chance for premature termination of a green duration increases when detector lengths are shortened, and a detector length of longer than 80 fit can effectively eliminate the premature termination. Using long detectors, however, results in longer dwell times and may reduce control efficiency. Lin and Shen [28] also indicated that the modeling of the vehicle-detector interactions should take into account the stochastic aspects of queuing in relation to detectors. The use of average characteristics of departure headway could result in underestimates or overestimates of the probability of premature termination of the green. AppendixA: Page ~

OCR for page 39
Later, Fin t29] evaluated the queue dissipation simulation models for analysis of presence-mode fiull- actuated signal control. The queue dissipation models used in the NET SIM program and the Value Iteration Process--Actuated Signals (V1PAS) program were evaluated. He indicated that both models were capable of producing realistic departures of queuing vehicles from the detector area. The models were rather weak, however, in representing other aspects of vehicle-detector interactions. A major weakness of the mode} used by NET SIM was that the simulated movements of queuing vehicles have little to do with the discharge times generated separately from a probability distribution. The weakness of V1PAS was that the University of Pittsburgh car-follow~ng mode} used in V1PAS did not provide a flexible model structure for calibration. Therefore, the outputs of the mode! could not be made to conform easily and simultaneously with observed departure, arrival, and dwell characteristics of queuing vehicles. hn 198S, Chang and Williams [30] investigated the assumption that independent vehicle arrivals at traffic signals, such as in the Poisson distnbution, have been widely used for modeling delay at urban intersections. The study introduced an effective yet economic approach to estimate the degree of correlation among among vehicles under given conditions and geometric characteristics. With the proposed technique, traffic professionals can easily determine if the existing delay formulas and other traffic simulation models based on the Poisson distribution are applicable. The presence of high variability in traffic simulation results often leads to concern about their reliability, and consequently precludes a rigorous evaluation of the target traffic systems performance under various control strategies. In 1990, Chang and Kanaan [3 1] presented the variability assess- ment for TRAF-NETSIM. The batch-means method, which allows the user to assess the variability of parameters, such as the average delay per vehicle, through a single relative long run, was introduced. This study provided a good contribution to traffic simulation users, given the large expenditures on computer simulation. ESTIMATION OF GREEN TIMES AND CYCLE LENGTH FOR TRAFFIC-ACTUATED CONTROL Traffic-related phenomena at a signalized intersection, such as lane capacity, delay and queue length are influenced by the phase times and cycle length. For traff~c-actuated control, green split and cycle length fluctuate with respect to the traffic demand. Consequently, it becomes desirable to predict the average green times and cycle length. In 1982, Lin [32] began to develop a mode! to estimate the average phase duration for full-actuated signals. The mode} was developed primarily on the basis of probabilistic interactions between traffic flows and the control system. He assumed the arrival at the upstream side of an intersection would be random' so the arrival pattern in each lane was represented by a Poisson distribution. Later, Lin and Mazbeyasna t33] developed delay models for semi-actuated and fi~-actuated controls that employ motion detectors and sequential phasing. These models were based on a modified version of Webster's formula. The modifications included the use of average cycle length, average green duration, and tvvo coefficients of sensitivity reflecting the degree of delay sensitivity to a given combination of traffic and control conditions. In 1990, Lin [34] proposed an Appendix A: Page 9

OCR for page 39
. improved method for estimating average cycle lengths and green intervals for semi-actuated signal operations as mentioned before. In 1994, Akcelik [35,36] proposed an analytical methodology for the estimation of green times and cycle time for vehicle-actuated signals based on the bunched exponential distribution of amval headway. The discussions in his papers were limited to the operation of a basic actuated controller that used passage detectors and a fixed gap time setting. Both fi~-actuated and sem~-actuated control cases were studied. A discussion of arrival headway distributions was presented since the estimation of arrival headways is fundamental to the modeling of actuated signal timings. The fonnulae were derived to estimate the green times and cycle tune based on the bunched exponential distribution of arrival headway. The bunched amval mode} was proposed by Cowan [37] and used extensively by Troutbeck [3S,39,40,4l,42,43] for estimating capacity and performance of roundabouts and other unsignal~zed intersections. The bunched arrival mode} considers that the bunched relationship between vehicles increases when the flow arrival rate increases. Since the bunched arrival mode! appears to be more representative of real-life amval patterns in general, Akc,elik used this arrival mode! for deriving various formulae for the analysis of traffic-actuated signal operations. The random amval mode! which uses negative exponential or shift negative exponential distribution of arrival headway can be derived as a special case of the bunched arrival mode} through simplifying assumptions about bunching characteristics of the arrival stream. The methods given in his papers provide essential information (average green times and cycle length) for predicting the performance characteristics (capacity, degree of saturation, delay, queue length and stopped rate) of intersections. DELAY MODELS FOR TRAFFIC-A CTUA1~D CON 1 HOL With the increase of computer software, the comparison of different traffic programs for pretimed and actuated controls became intuitively appealing. In ~ 974, Nemeth and Mekemson [44] compared the delay and filet consumption between deterministic Signal Operations Analysis Package (SOAP79) and the microscopic and stochastic Network Simulation Mode! (BETSY for pretimed and actuated controls. They indicated that in terms of delay prediction, SOAP79 and NETSIM were found to be entirely compatible except for the difference in clelay definitions. In 198S, Akcelik [45] evaluated the 1985 HCM [~] delay formula for signalized intersections. He stated that the HCM formula predicted higher delays for oversaturated conditions. An alternative equation to the HCM formula was proposed. This formula gave values close to the HCM formula for degrees of saturation less than I.0, and at the same time, is similar to the Australian, Canadian, and TRANSYT formulas in producing a delay curve asymptotic to the deterministic delay line for a degree of saturation greater than I.0. The signalized intersection methodology presented in the 1985 HCM [~] introduced a new delay model. Lin [46] evaluated the delay estimated by the HCM with delay observed in a field study in 1989. Some inconsistency existed in the delay estimation between the HCM results and the field AppendixA: Page 10

OCR for page 39
observation. He suggested improving the progression adjustment in the HCM procedure and using a reliable method to estimate average cycle lengths and green durations for traffic-actuated signal operations. ~ 1989, Hagen and Courage [47] compared the HCM [~] delay computations with those performed by S0APS4 and TRANSYT-7F Release 5. The paper focused on the elect of the degree of saturation, the peak-hour factor, and the period length on delay computations and the treatment of leer burns opposed by oncoming traffic. They indicated that all ofthe models agreed closely at volume level below the saturation point. When conditions become oversaturated, the models diverged; however, they could be made to agree by the proper choice of parameters. The computed saturation flow rates for led turns opposed by oncoming traffic also agreed closely. However, the treatment of protected plus permitted left turns produced substantial differences. It was concluded that neither SOAP nor HCM treats this case adequately. A delay mode} was recommended in the HCM tI] for level-of-serv~ce analysis at signalized intersections. The use of this mode} for the evaluation of traffic-actuated signal operations required a knowledge of the average cycle lengths and green intervals associated with the signal operation being analyzed. Since the method suggested in the HCM to estimate delay of traffic-actuated signal operations was not reliable, Lin [34] proposed an improved method for estimating average cycle lengths and green Intervals in 1990. The method was limited to sem~-actuated signal operations. Din stated that this method, which was developed through analytical modeling and computer simulation, was .~fficient~v s~mnIe and reliable Realistic examples were used to illustrate the application of the method. In 1993, IN Rouphail and Akcelik [48] presented an approach for estimating overflow delays for lane groups under traffic-actuated control using the 1985 HCM delay mode} format. The average cycle lengths and green times used in the delay mode! were obt-a~ned Dom a cycle-by-cycle simulation model. This study was limited, however, to two phase single lane conditions. The results indicated that the average cycle and green times are very much related to the controller settings such as mrnimum and maximum green times, cycle lengths, and extension times, with longer extension times producing higher cycle length. It was found that overflow delay increases with longer extension times. Further, by applying the 1985 HCM delay formula to the simulated signal settings, the com- puted delays were much higher. This implies the need for separate calibration of the second delay term to account for the actuated control effects. SIGNALIZED INTERSECTION CAPACITY MODELS FOR TRAFFIC-ACTUATED CONTROL Intersection capacity analysis is essential for measurements of most traffic control effectiveness. The first U.S. HCM (1950) contained a chapter for estimating the capacities of signalized intersections. Numerous studies were undertaken to evaluate the different aspects of signalized intersections and many capacity methods were developed. In 1983, May, Gedizlioglu and Tai t491 began the evalua- tion of eight available methods for capacity and traffic-perfo~mance analysis at signalized intersections AppendixA: Page 11

OCR for page 39
including pretimed and actuated controls. The eight methods included the U. S. Highway Capacity Manual method (1965), the British method (1966), the Swedish method (1977), the Transportation Research Board (TRY) Circular 212 planning method (1980), the TRB Circular 212 operations and design method (19SO), the Australian method (198I), the National Cooperative Highway Research Program (NCHRP) planning method (1982), and the NCHRP operations method (1982). They concluded that the NCHRP operations method and the Australian method were found to be the most cost-effective methods. In 1991, Prevedouros [50] studied the traffic measurements and capacity analysis for actuated signal operations. In chapter 9 of the 1985 HCM, he thought it was not appropriate to treat the pretimed and actuated controls identically, especially concerning the estimation of capacity and performance of existing Intersections. The study verified the fact that intersections with traffic-actuated signal controls demand special treatment when estimating their capacity and performance. The main sources of error and their potential impacts were presented. He developed a comprehensive data collection and analysis methodology which complemented the procedure in the 1985 HCM. The major findings were: I) the g/C ratio should be estimated Tom a consistent and complete series of field measure- ments if accurate results are to be obtained; 2) the true peak hour and peak ~ 5-minute period can be identified only when cycle-by-cycle data are used; 3) small variations of the peak period of analysis around the standard length of ~ 5 minutes have a negligible impact on the delay estimation. OVERVIEW AND EVALUATION OF EVIPAS EV1PAS is an opti~zation/simulation mode! for actuated controlled, isolated intersections. It is capable of analyzing and determining the optimal settings of timing parameters for a wide range of geometric configurations, detector layouts, and almost any phasing pattern available in a single or dual-ring NEMA and Type 170 controllers. It wait generate the optimized timing settings for controllers ranging from pretimed to volume-density actuated controllers. The optimum settings of timing parameters include minimum green time, maximum green time, unit extension time, minimum gap, time before reduction, time to reduce, added initial, and maximum initial for each phase. The value of optunal timing is defined as the timing setting which results in the minimum "total cost." The mode! allows the user to define the total cost to include a variety of measures of effectiveness, such as delay, filet consumption, depreciation, other vehicle costs and emissions. The POPPAS mode! allows for two modes of operation. In its optimization mode, the mode! is used to obtain optimal timing settings by a multivariate gradient search optimization module and an event-based intersection m~crosimulation. In the simulation mode, EV1PAS allows the evaluation of a prespecified signal plan just by m~crosimulation. The input data for the EVIPAS mode! is in text fo~.~at, organized into two separate input files. To facilitate the use of EV1PAS, a more fiiendly input processor called EZV1PAS [5 I], was developed. The purpose of EZV1PAS is to ease the data input effort and eliminate coding errors. The input data is provided by EZV1PAS to the EVE?AS mode! via two text files: a DAT file and a RUN file. Each file consists of a number of record groups. The data on the DAT file consists mostly of geometric AppendixA: Page 12

OCR for page 39
data and other data elements which typically remain fixed throughout multiple EVIPAS runs to analyze various operational features of the study site. The RUN file consists mostly of the signalization, timing and optimization data. In DAT files, separate screens are provided for: Identification and Default Ovemdes, one screen. Approach Information, one screen. Lane and Detector Information, one screen per approach. Signal System Setup, one screen. Signal Phasing Definition, one screen. Period Definition, one screen. Traffic Volume, one screen per period. Pedestrian Flow, one screen. In RUN files, separate screens are provided for: Run Control, one screen. Cost and Emission Parameters, one screen. Signal Timing Start Values, one screen. Optimization Flags, one screen. Lower and Upper Bounds for Optimized Parameter, one screen per EV]PAS signal section. . The EVIPAS model output can be classified into four groups consisting of I) input data echo tables, 2) summary of vehicle generation, 3) summary of optimization progress, and 4) summary of performance output. For the capacity and level of service oftraffic-actuated control, the performance outputs are primarily concerned. The summed of performance output is a set oftables reporting the final results of the EV1PAS simulation/optim~zation run and consists of I) summary of delays table, 2) signal performance summary table, 3) summary of final signal setting table, and 4) sumunary of final costs table. For this study, summary of delays table and signal performance summary table are more concerned. The summary of delay table provides delay statistics for the intersection and for each approach and lane. The signal performance summary table shows the average phase length and average cycle length. All above delay measures, average phase length, and average cycle length are based on the m~crosimulation results. Since EV]PAS is a simulation-based model, it becomes necessary to compare the simulation results from EV1PAS with other simulation programs. NET SIM is an appropriate candidate for this purpose. The current version of NETSIM produces very detailed tables of several performance measures. It does not, however, provide sufficient information on the operation of the controller itself in the standard output tables. To obtain this information, it was necessary to develop a postprocessor to extract the operational data from special files used to support the animated graphics features of NETSIM. The actuated-controller data for each second of operation are recorded and stored in a text file that is given a file name extension of ".F45" by NETSIM. Through the CNVX45 program, the .F45 file can be converted to a readable .X45 text file. A postprocessor was developed to read the .X45 file and produce a summary of the operation. The postprocessor has been called AppendixA: Page 13

OCR for page 39
"NETCOP" for "NETSIM Controller Operation Postprocessor." It can provide phase-specific infonnation such as percent skipped, percent gapout, percent maxout, average phase time, average cycle length, and adjusted cycle length. The "Adjusted Cycle Length" is computed by subtracting the number of seconds of awed (i.e., the time during which no demand was registered on any phase) from the total number of seconds simulated before dividing by the number of cycles. Although the simulation techniques used in EV]PAS and N.ETSIM maY differ in some decrees ~ , ~ . - ~ theoretically, the phase time estimations tor the same traffic conditions, geometric con~gurat~ons and actuated timing settings should be close. Thus, the evaluation of EV1PAS was performed by comparing its simulated phase times with the phase times produced by NET SIM. A total of 15 HEM hypothetical traffic-actuated examples were used. These examples cover both two-phase and multi- phase actuated operation. Results Dom the phase time comparison between EV]PAS and NETSIM are shown ~ Figure Art. 70 U 60 C) ._ al 40: lo 30 ._ En 2~ cc 10 50 _ ,, ,, , ,,,,,,, , , ,,,,,,, , ,. _ __ _.-.- _ -- 7 2 R = 0.96 icy _ ~ ~ ,~ a~' a3~- EW''-'''''-----'''':----''''''''--'''-''''''''''''''''''''''''''''''-'''--'---'''''''''''''''''''''''-- o 0 10 20 ~3 ~ 50 60 70 NETSIM Simulated Phase Time (see) Figure AM Phase time comparison between EV1PAS and NETSIM A simple regression analysis between the above two simulated phase times was performed and a 0.96 coefficient of determination, R2, was achieved. As expected, the simulated phase times Dom EV1PAS are very close to those Dom NET SIM simulations. This suggests that, when estimating phase times, the EVIPAS mode} is as effective as the NET SIM model. AppendixA: Page 14

OCR for page 39
APPENDIX A REFERENCES 1. Transportation Research Board, "Highway Capacity Manual," Special Report 209, Washing tonD.C., 1985. 2. Orcutt, F. L., "A primer for Traffic Signal Selection, " Public Works-City, County and State. March 1975, pp. 76-80. 3. National Electrical Manufactures Association, ['Traffic Control System," Standards Publica tion No. TS1, 1976. Vodrazka' W. C., C. E. Lee and H. E. Haenel, "Traffic Delay and Warrants for Control Devices, " HRB Record 366, 1971, pp. 79-91. 5. Gerlough, D. L. and F. A. Wagner, "Improved Criteria for Traffic Signal at Individual Inter sections,'l NCHRP Report 32, 1976 6. Bang, K. and L. Nilsson, "Traffic Signal Control with Long Loop detectors," Traffic Engi neering and Control, Vol. 11, No. 11, March 1970, pp. 525-527. 7. Cribbins, P. E. and C. A. Meyer, "Choosing Detectors and Placing Them Right," Traffic Engi neenng, Vol. 45, No. 1, January 1975, pp. 15-18. 8. Webster, F. V., "Traffic Signal Settings," Road Research Paper No. 39, Scientific and Indus trial Research, HMSO, London, 1958. 9. Miller, A. J., "Settings for Fixed-Cycle Traffic Signals," Operations Research Quarterly, Vol. 14, No. 4, 1963, pp. 373-386. 10. Newell, G. F., "Properties of Vehicle-Actuated Signals: I. One-Way Streets," Transportation Science, Vol. 3, No. 1, Feb. 1969, pp. 30-52. 11. Newell, G. F. and E. E. Osuma, "Properties of Vehicle-Actuated Signals: II Two-Way Streets," Transportation Science, Vol. 3, No. 2, May 1969, pp.99-125. Staunton, M. M., "Vehicle Actuated Signal Controls for Isolated Locations," AN FORAS FORBARTHA, The National Institute for Physical Planning and Construction Research, No.RT- 159, Dublin, Ireland, Sept. 1976. 13. Bang, K. L., "Optimal Control of Isolated Traffic Signals," Traffic Engineering and Control, July 1976, pp. 288-292. 14. Tarnoff, P. and P. S. Parsonson, llSelecting Traffic Signal Control at Individual Intersections,ll NCHRP Report 233, Transportation Research Board, National Research Council, Washington D.C., June 1981. AppendixA:Page15

OCR for page 39
15. L`in, Feng-Bor, "Optimal Timing Setting and Detector Lengths of Presence Mode Full Actuated Control," Transportation Research Record 1010, TRB, National Research Council, Washington D.C., 1985, pp. 37-45. 16. Kell, I. H. and T. I. Fullerton, "Manual of Traffic Signal Design," Prentice-Hall, Englewood Cliffs,N.J., 1982. 17. Bullen, A. G. R., N. Hummon, T. Bryer and R. Nekmat, "EV1PAS:A Computer Mode! for the Optimal Design of a Vehicle-Actuated Traffic Signal," Transportation Research Record ~ ~ 14, TRB, National Research Council, Washington D.C., 1987, pp. 103-! 10. IS. Messer, C. I. and M. Chang, "Traffic Operation of Basic Traffic-Actuated Control Systems at Diamond Interchanges," Transportation Research Record ~ ~ 14, TRB, National Research Council, Washington D.C., 1987, pp. 54-62. 19. Courage, K. G. and I. Z. Luh, "Development of Guidelines for Implementing Computerized Timing Design at Traffic Actuated Signals, Final Report Volume I," Isolated Intersection Implementation, Transportation Research Center, University of Florida, February 1989. 20. BuDen A. G. R. "The Ejects of Actuated Signal Settings and Detector Placement on Vehicle Delay," Presented at Transportation Research Board 68th Annual Meeting at Washington, D.C., 1989. 21. Kell, I. H. and T. I. Fullerton, "Manual of Traffic Signal Design, 2nd Edition," Institute of Transportation Engineers, Prentice-Hall, Englewood Cliffs, N. I., 1991. 22. Bonneson, I. A. and P. T. McCoy, "A Methodology for Evaluating Traffic Detector Designs," Transportation Research Record 1421, TRB, National Research Council, Washington D.C.,1994, pp 76-~. 23. Jovanis P. P. and I. A. Gregor, "Coordination of Actuated Arterial Traffic Signal System," Journal of Transportation Engineering, Vol. ~ 12, No. 4, July 1986. 24. Courage, K. G. and C. E. Wallace, "Development of Guidelines for Implementing Compu terized Timing Design at Traffic Actuated Signals, Final Report Volume 2," Arterial System Implementation, Transportation Research Center, University of Florida, February 1989. 25. Chang, E. Chin-Ping and I. Koothrappally, "Field Verification of Coordinated Actuated Control," TRB 73rd Annual Meeting, January 1994, Washington D.C. 26. Lull, I. Z. and K. G. Courage, "Late-N~ght Traffic Signal Control Strategies for Arterial Sys tems," Transportation Research Record 1287, TRB, National Research Council, Washington D.C., 1990-1991, pp. 205-211. 27. Lin, Feng-Bor and M. C. Percy, "Vehicle-Detector Intersections and Analysis of Traffic Actuated Signal Controls," Transportation Research Record 971, TRB, National Research Council, Washington D. C., June ~ 984, pp. ~ ~ 2- ~ 20. Appendix A: Page 16

OCR for page 39
28. Lin, Feng-Bor and S. Chen, "Relationships between Queuing Flows and Presence Detectors," ITE Journal Vol. 55, No. 8, August 1985. 29. Lin, Feng-Bor, "Evaluation of Queue Dissipation Simulation Models for Analysis of Presence-Mode Fud-Actuated Signal Control'" Transportation Research Record 1005, TRB, National Research Council, Washington D.C., 1984, pp. 46-54. 30. Chang, Gang-Len and J. C. Williams, "Estimation of Independence of Vehicle Arrival at Signalized Intersections: A Modeding Methodology," Transportation Research Record ~ 194, TRB, National Research Council, Washington D.C., INS, pp. 42-47. Chang, Gang-Len and A. Kanaan, "Variability Assessment for TRAF-NETSIM," Journal of Transportation Engineering, Vol. ~ ~ 6, No. 5, September/October, ~ 990. 32. Lin, Feng-Bor, "Estimation of Average Phase durations for Full-Actuated Signals," Transportation Research Record 881, TRB, National Research Council, Washington D.C., 1982, pp. 65-72. 33. Lin, Feng-Bor and F. Mazdeyasna, "Delay Model of Traffic-Actuated Signal Controls," Transportation Research Record 905, TRB, National Research Council, Washington D.C., 1983-1984, pp. 33-38. 34. Lin, Feng-Bor, "Estimating Average Cycle Lengths and Green Intervals of Sem~actuated Signal Operations for Level-of-Service Analysis," Transportation Research Record 1287, TRB, National Research Council, Washington D.C., 1990-1991, pp. ~ 19-128. 35. Ak~elik, R., "Analysis of Vehicle-Actuated Signal Operations," Australian Road Research Board. Working Paper WD TE 93/007, 1993. 36. Ak',celik, R., "Estimation of Green Times and Cycle Time for Vehicle-Actuated Signals," Presented at Transportation Research Board 73rd Annual Meeting at Washington D.C., 1994. 3 7. Cowan, R. I., "Useful Headway Models," Transportation Research 9 (6), ~ 975, pp. 3 7 ~ -3 75. 38. Troutbeck, R. I., "Does Gap Acceptance Theory Adequately Predict the Capacity of a Roundabout," Proc. 12th ARRB Conf. 12 (4), 1984, pp. 62-75. 39. Troutbeck, R. J., "Average Delay at an Unsignalized Intersection with Two Major Stream Each Having a Dichotomised Headway Distribution'', Transportation Science 20 (4), 1986, pp. 272-286. 40. Troutbeck, R. J., "Current and Future Australian Practices for the Design of Unsignalized Intersections," In: Intersections Without Traffic Signals (Edited by W. Brilon), Proceedings of an International Workshop (Bochum, West Germany), Springer-Veriag, Berlin, INS, pp. I_19 Appendix A: Page 17

OCR for page 39
41. Troutbeck, R. I., "Evaluating the Performance of a Roundabout," Special Report SR 45, Australian Road Research Board, ~ 989. 42. 43. 44. 45. 46. Troutbeck, R. I., "Roundabout Capacity and the Associated Delay" In: Transportation and Traffic Theory (Edited by M. Koshi), Proceedings of the Eleventh International Symposium on Transportation and Traffic Theory, (Yokohama, Japan), Elsevier, New York, ~ 990, pp. 39-57. Troutbeck R I., "Recent Australian Unsignalized Intersection Research and Practices," In: Intersections Without Traffic Signals IT (Edited by W. BnIon), Proceedings of an International Workshop (Bochum, West Germany), Springer-VerIag, Berlin, 1991, pp. 239- 257. Nemeth, Z. A. and I. R. Mekemson, "Comparison of SOAP and NETSIM: Pretimed and Actuated Signal Controls," Transportation Research Record 905, TRB, National Research Council, Washington D.C., ~983-~984, pp. 84-89. Ak~elik, R., "The Highway Capacity Manual Delay Formula for Signalized Intersections," ITE Journal 58 (3), march, INS, pp. 23-27. Lin, Feng-Bor, "Applications of 1985 Highway Capacity Manual for Estimating Delay at Signalized Intersections," Transportation Research Record 1225, TRB, National Research Council, Washington D.C., 1989, pp. IS-23. 47. Hagen, L. T. and K. G. Courage, "Comparison of Macroscopic Models for Signalized Intersection Analysis," Transportation Research Record 1225, TRB, National Research Council, Washington D.C., 1989, pp. 33-44. 48. 49. Li, Jing, N. M. Rouphai} and R. Ak~elik., "Overflow Delay Estimation for a Simple Intersection with Fully-actuated Signal Control," Presented at Transportation Research Board 73rd Annual Meeting at Washington D.C., 1994. May, A. D. and E. Gedizlioglu and L. Tai, "Comparative Analysis of Signalized Intersection Capacity Methods," Transportation Research Record 905, TRB, National Research Council, Washington D. C., ~ 983 - ~ 984, pp. ~ ~ 8- ~ 27. 50. Prevedouros, P. D, "Actuated Signalization: Traffic Measurements and Capacity Analysis," ITE 1991 Compendium ofTechnicalPapers. 51. EZV1PAS I.0 User Guide, Viggen Corporation, 1993. Appendix A: Page 18