Transportation Research Record: Journal of the Transportation Research Board

The Highway Capacity Manual (HCM) did not originally contain an arterial planning methodology, but the state of Florida perceived an acute need for one in order to meet its obligations for growth management. Such a methodology was developed, documented, and implemented throughout the state. The methodology was firmly based upon the HCM detailed analysis procedure but had assumptions and defaults that allowed a range of planners to use it effectively. The basic construct was to provide tables of design volumes v and annual average daily traffic that could be achieved for various levels of service and signal spacings for the default parameter values. Two computational tools, implemented in the form of spreadsheet templates as being the most user-friendly and open approach, were provided to users so that they could modify the default settings or specify particular signal spacings and such. The methodology has been in widespread use in Florida, it has served as the basis for the arterial planning technique incorporated into the last update to the HCM, and other states have requested information. The experience with the implementation, including training needs and common user questions, is recounted.

An arterial planning methodology has been in use in Florida for several years and influences decisions in developments of regional impact and other analyses. It is incorporated into the Florida Level of Service Manual. Experience has indicated that analysts using the default values are concerned that the average travel speeds are underestimated by 10 km/h or more. Given the known limits of the data underlying the arterial running time and running speed default estimates in the Highway Capacity Manual (HCM), it was decided to conduct an extensive simulation analysis using a well-established model (NETSIM, now part of CORSIM). The findings are that (a) simulation-derived running speed estimates are at least some 10 km/h higher than the HCM default values, (b) moreover, the running speed estimates generally do not decrease as sharply as the HCM default values as signal density increases, and (c) unlike the HCM default values, the simulation-derived estimates are sensitive to increased demand volume. In addition, the derived values clearly reflect the through vehicles, consistent with the original HCM intent.

The objective of the research described in this paper was to develop the basis for revised operational analysis procedures for transportation facilities with pedestrian users where flow is not interrupted by traffic control devices. This paper contains both new and revised level-of-service (LOS) tables for analyzing various types of uninterrupted-flow pedestrian facilities. It details the results of a review and synthesis of American and international literature as part of a Federal Highway Administration study of pedestrian and bicycle facilities conducted by North Carolina State University between 1995 and 1998. The year 2000 edition of the U.S. Highway Capacity Manual (HCM) will incorporate, either directly or indirectly, most of the revised and synthesized information described in this paper. The research team recommends that the HCM revise the pedestrian walkway level of service to reflect the current literature. Proposed revisions include a decrease in the threshold for LOS A from 130 to 60 ft²/ped (12 to 5.6 m²/ped) and a change in the capacity level from 6 to 8 ft²/ped (0.6 to 0.75 m²/ped). The incorporation of a simplified walkway platoon level of service is recommended. In addition, given the different expectations of acceptable levels of platooning in transportation terminals, different LOS criteria are recommended for these facilities. A method of analyzing pedestrians on shared pedestrian-bicycle paths has also been identified. Although developed in the Netherlands, the procedure has been validated on at least one trail in the United States and is therefore recommended for inclusion in the HCM.

The objective of the research described in this paper was to develop the basis for revised operational analysis procedures for transportation facilities with pedestrian users where flow is interrupted by traffic control devices. The paper commences with some background information on pedestrian walking speeds at signalized crossings and on pedestrian noncompliance at these locations. Then both new and revised level-of-service (LOS) tables are provided for analyzing various types of interrupted-flow pedestrian facilities. Results are detailed of a review and synthesis of American and international literature as part of a Federal Highway Administration study of pedestrian and bicycle facilities conducted by North Carolina State University between 1995 and 1998. The year 2000 edition of the U.S. Highway Capacity Manual (HCM) will incorporate, either directly or indirectly, most of the revised and synthesized information described in this paper. The research team recommends that the HCM include some background information that will be helpful for analysts timing signals and performing other operations. The authors of this paper recommend assumed crosswalk walking speeds of 1.2 m/s for most areas and 1.0 m/s for crosswalks serving large numbers of older pedestrians. Another important recommendation was to use delay to pedestrians as the basis for level of service at signalized and unsignalized street crossings. The recommended threshold for LOS F at signalized crossings is 60 s or more of delay per pedestrian, whereas at unsignalized crossings it is 45 s or more of delay per pedestrian.

Signal progression plays an important role in controlling arrival pattern and delay at coordinated signalized intersections. In the Highway Capacity Manual (HCM), one random and five nonrandom arrival types are identified. The delays for nonrandom arrival cases (platoon arrivals) are computed by applying progression adjustment factors (PFs) to the delay for the random arrival case. However, HCM-estimated delays are found to be significantly different from field-measured delays for Arrival Types 1, 4, and 6. For Arrival Type 2 the HCM estimations are acceptable, although the deviations are still relatively high. For Arrival Types 3 and 5 the HCM estimates the delays accurately. These findings suggest that the HCM PFs need to be revised. A new set of PFs was developed using the arrival-based delay models. A relationship between the effective green ratio, g/C, and the new PFs was established for each arrival type. With the proposed PF values, the HCM approach would be more accurate than it is with the current PF values. For Arrival Type 1, the new PF values are higher than the current HCM values for g/C ⩽ 0.50 and lower for g/C > 0.50. For Arrival Types 2 and 4, the new PF values are lower than those of HCM, as expected, because the HCM overestimates the delays for both Arrival Types 2 and 4. For Arrival Type 6 the new PF values are lower for g/C ⩽ 0.40; otherwise they are higher. A new progression adjustment table similar to Table 9-13 of the HCM is created. It is recommended that the new PF values be adopted by the HCM since they enhance the accuracy of delay estimation.

When vehicles in a queue start crossing a stop line (or any other reference line) at a signalized intersection after the signal turns green, the entering headway (discharge headway) is the time that elapses between consecutive vehicles. This headway is a factor in determining intersection capacity. The current study aims at analyzing the entering headway at selected signalized intersections in Riyadh (the capital of Saudi Arabia). It is unlikely that driver behavior is similar in different countries, and therefore it is unwise to use the values for headway times from other countries to fit traffic conditions in Saudi Arabia. To date, no research of this type has been conducted in Saudi Arabia. This study attempts to investigate the entering headway at intersections in Riyadh and to compare the results with corresponding results from international research. Eight signalized intersections (with two-lane and three-lane approaches) along eight major streets in Riyadh were selected for study. A total of 720 single-lane traffic platoons entering the intersections were observed. It was found that the average entering headways for Vehicles 1 through 15 at two-lane intersections in Riyadh after the light turned green were 3.23, 2.41, 2.29, 2.1, 2.02, 1.84, 1.72, 1.64, 1.53, 1.68, 1.58, 1.31, 1.13, 1.19, and 1.01 s. For the three-lane sites the headways for Vehicles 1 through 13 were as follows: 3.37, 2.23, 2.03, 1.92, 1.81, 1.81, 1.72, 1.69, 1.6, 1.48, 1.37, 1.33, and 1.46 s. The study found that the saturation flow rates were 2,293 and 2,195 vehicles/hour of green per lane for two-lane and three-lane sites, respectively. These intersection approaches appear to have higher saturation flow rates than those in previous studies from other countries.

Many traffic engineers nationwide are reluctant to implement lagging left-turn phasing plans along arterial streets, even though signal progression can sometimes be significantly improved. An opinion frequently expressed by traffic engineers is that lagging left turns are contrary to driver expectancy and cause an increase in start-up lost time. To prevent the “left-turn trap” in which drivers making left turns may be caught unaware of the extended opposing lagging phase, many traffic engineers do not allow permitted left turns at an intersection with lagging phasing. This study focuses only on protected left-turn phasing in which the trap does not occur. Saturation flow rates, start-up lost times, fourth-vehicle crossing times, and accident rates are compared between protected-only leading and lagging phasings. All four measures were found to be better for locations with lagging phasing than for those with leading phasing, although neither saturation flow rates nor accident rates were found to be significantly better at the 95 percent confidence level. Although further studies are needed, these findings suggest that a mix of leading and lagging protected-only phasing may be used to gain signal coordination advantages without compromising safety or capacity.

Many roundabouts have been built in the Netherlands over the last 15 years. A much-debated issue has been whether to give slow traffic, especially cyclists, priority. Recently, a general recommendation was made: priority for slow traffic inside built-up areas and the reverse for rural roundabouts. Because the implementation of this recommendation could result in a large number of roundabouts with priority for slow traffic, there was a need to develop a method that determined the capacity and delay, taking the influence of slow traffic fully into account. In this study such a model was developed. It is an analytical model that combines parts of existing analytical submodels, with extensions. Motorized traffic entering a roundabout is confronted with three conflicting streams: the well-known circulating stream of vehicles on the roundabout, the parallel slow traffic blocking the entry, and the circulating stream of slow traffic at the next downstream exit that blocks the outgoing motorized vehicles and as a consequence can also block the entry. The model was calibrated successfully with observations from two one-lane saturated roundabouts. With the model not only the capacity per entry can be calculated but also the capacity of the total roundabout, given a relative origin-destination matrix of fast and slow traffic.

The purpose of this work is to investigate the relationship between speed and flow within congestion, that is, the lower portion of the standard curve, and in particular to identify an equation to describe this portion of the relationship. Data were obtained for Toronto, Ontario, from the Gardiner Expressway RESCU System and the Highway 401 COMPASS System. The data were first screened to remove dubious data points and observations that were from transitions between congested and uncongested conditions. They were then analyzed using regression analysis for four types of function: quadratic, cubic, exponential, and power. The principal findings related to the shape of the speed-flow curve within congestion are as follows. It is important to utilize data from a full range of flows in order to fit a curve to represent the congested part of the speed-flow curve. To obtain this range, data from several sites may be needed. Despite the combination of sites, there is often what might be termed a “data gap” between easily available congested data (up to perhaps 1,800 vphpl) and flow rates for queue discharge flow. When data for that gap are available, the speeds under congested conditions seem to increase fairly rapidly at high congested flows. There appears to be a difference in the speed-flow relationship among different freeways and under construction conditions on one of them. If there are different curves for different freeways (or free-flow speeds) for the top half of the speedflow relationship, it seems reasonable that there might also be for the bottom half. However, it also appears to be the case that downstream conditions such as construction can affect the speed-flow curve itself at an upstream location. This makes sense in that operations within the queue are governed by the downstream queue discharge, but it is a change from what one expects for the other two segments of the speedflow curve.

The National Highway System Designation Act of 1995 repealed the national maximum speed limit. As a result, many of the states have increased posted speed limits to 105 km/h (65 mph) for select multilane highways. This study evaluates the distribution of the speed and studies the influence of site characteristics, light conditions, and type of vehicle on the field-measured free-flow speed. The suitability of the Highway Capacity Manual multilane highway rules of thumb for freeflow speed to both an 89- and a 105-km/h (55- and 65-mph) posted speed limit condition is also discussed.

Since about 1990 the quality of road traffic operation on Dutch freeways is no longer expressed as a level of service, but in the percentage of vehicles that experience congestion and in terms of their delay. The criterion for an optimal congestion percentage refers to the overall value for all working days of a year. This overall percentage is determined by means of a deterministic queueing model that simulates a year’s traffic operation. A simplified version of this model is presented in this paper. It takes into account systematic and random fluctuations in traffic demand and roadway capacity. Demand is a function of the month of the year, the day of the week, and the hour of the day, and it has a random fluctuation. Capacity changes because of light and weather conditions and also has a random component. The extent to which several sources of variation contribute to the overall percentage of vehicles experiencing congestion is analyzed as well as the total delay and the delay per vehicle involved. Moreover, the day-to-day variation of several congestion characteristics and their dependencies are discussed.

Presented in this paper are the results of research on passing activity on two-lane highways in Spain. The data were gathered on three highways in the province of Madrid, Spain (M-111, M-600, and M-607). Measurements were carried out in 3-h periods, on plain or slightly sloping ground, and for daily or weekly peak periods. The average daily traffic and rate of heavy vehicles in 1991 were, sorted chronologically, 9,500 (6.4 percent) on the M-607, 9,800 (5.1 percent) on the M-600, and 8,100 (8.6 percent) on the M-111. Passing maneuvers were accounted for by examining the differences in order of vehicles at entry and exit points for each road section. A total of 780 passing maneuvers were detected. Passing rates were studied, in particular the variation of passes made with flow rates, both in the direction of passing and in the opposing direction. Maximum passing rates are suggested for each opposing flow rate, and it is indicated that passing is most effective for balanced flows of between 500 and 700 veh/h.

Simulation is often used to address issues that cannot be effectively resolved using the Highway Capacity Manual (HCM) or other analytical procedures. However, guidelines are not available for appropriate application of simulation as a complement or substitute for HCM procedures. The purpose of this paper is threefold: (a) to provide an overview of simulation models in a context consistent with facilities and situations addressed by the HCM, (b) to present a framework for the appropriate selection of simulation models when procedures of the HCM are not adequate, and (c) to provide brief examples demonstrating application of the proposed framework. The paper is divided into five sections. The first section presents simulation concepts and definitions. The second presents a suggested framework for application of simulation within an HCM context. The third and fourth sections apply this framework to the evaluation of uninterrupted- and interrupted-flow facilities, respectively. In these, appropriate application of the proposed framework for integration of simulation into an HCM-type analysis framework is presented. Last, the relationship between the methodological framework presented in this paper and the upcoming HCM 2000 is discussed.

In high-occupancy-vehicle (HOV) facility design, there are concerns about the impacts of merging and weaving movements at the ingress and egress (I/E) sections of a barrier-separated freeway HOV lane. Weaves and merges can disturb traffic flow and cause freeway capacity reduction at HOV-lane I/E sections. Such capacity reductions can cause bottlenecks that could eventually reduce the benefit of an HOV lane. An impact analysis regarding capacity reduction is necessary, which can improve HOV-lane operations by leading to reasonable design options that minimize the capacity reduction. This study developed a method for estimating the capacity reductions at freeway HOV-lane I/E sections under various geometric and operating conditions, which is an effective tool for HOV facility planning, design, and operations analysis. The study first identified the major factors that affect freeway HOV-lane I/E section capacity. Then it explored the relationship between capacity reduction and the impact factors on the basis of a database developed from field collection and a freeway simulation model. Finally, the study developed an applicable equation for estimating capacity reduction using nonlinear regression with a third-order polynomial format. The equation was tested through a case study, the results of which indicate that the equation can yield reasonable results, satisfying the 95 percent confidence interval, in capacity reduction estimation.

Presented in this paper are the car-following methods and algorithms of the NETSIM, INTRAS, FRESIM, CARSIM, and INTELSIM models. Moreover, the car-following performance of these models is compared with the field data. NETSIM, INTRAS, FRESIM, and CARSIM car-following models first move the leader and then update the follower in one simulation time step. Because of this approach, these car-following models cannot be used to command vehicles in real-time intelligent transportation systems applications. Moreover, brake reaction times are limited by the simulation time step because of this method of updating the vehicles. INTELSIM was developed to overcome these deficiencies. INTELSIM moves vehicles simultaneously and produces solutions for a continuous time frame. INTELSIM produced the best agreement with the field data and required the least amount of calibration effort.

Recent German research has suggested that there may be two qualitatively different types of congested flow, referred to as “traffic jams” and “synchronized flow.” The possible existence of synchronized flow raises a theoretical issue, since it is reported to be characterized by randomly sloped flow-density “transferences” (lines connecting adjacent points in the time series), and this may imply that there is no consistent wave direction in such flow. An analysis of flow-occupancy data from San Diego suggests that on the basis of the appearance of the data alone, two types of congested flow may be identified. These appear to be similar to, but not identical with, the two types reported in the German work; in particular, a region with positively sloped flow transferences was found. Further consideration of the data and the mechanics of congested flow suggests, however, that these two types of flow do not represent qualitatively different types of traffic behavior but rather are the result of random variations in average time gaps in the traffic stream and the relationship among speed, flow, density, and average time gap. In particular, predominantly positively sloped flow-concentration transferences do not necessarily imply that waves move downstream; rather they imply that, for data taken over short time intervals (30 s or 1 min), transferences are not a good indicator of kinematic wave speed.

The cellular automata (CA) microsimulation of pedestrians is a particlehopping model in which a set of local rules prescribe the behavior of entities within local neighborhoods of cells. CA microsimulation has emerged as a tool for simulating traffic flow and modeling transportation networks. Pedestrian flow is inherently more complex than vehicular flow, and simulation models that are used for emulating vehicular traffic are not directly applicable to modeling pedestrian movements. In previous work the authors demonstrated that unidirectional pedestrian flow patterns consistent with well-established fundamental properties could be generated with CA microsimulation. This paper expands upon the previous effort and presents a CA microsimulation model and emergent fundamental flows for a bidirectional pedestrian walkway. Simulation experiments indicate that the basic model is applicable to walkways of various lengths and widths and across different directional shares of pedestrian movements.

This paper is concerned with the problem of on-line monitoring of a traffic network. The problem was motivated by the need of real-time dynamic traffic assignment systems to maintain an internal representation of the traffic network consistent with that of the actual network. Potential error sources that can cause inconsistencies between the internal and the actual state of the network are identified and classified. The issues of observability and state variable definition of traffic networks are discussed. A framework for a monitoring system is sketched that aims to maintain a representation of the state of a network consistent with that of the actual network by applying on-line and off-line adjustments. Some solution approaches and computational results are presented for the main modules of the framework.

A gas–kinetic traffic flow model is presented that describes the traffic flow operations of heterogeneous traffic on specific lanes of a motorway. To realistically portray the multilane dynamics of traffic flow, traffic is described as a collection of vehicle platoons that consist of an unconstrained vehicle followed by a nonnegative number of constrained followers. It is shown that the gas–kinetic dynamics of this unconstrained and constrained traffic are governed by continuum and noncontinuum processes. The continuum processes describe smooth changes in the mesoscopic traffic flow variables caused by motions in the phase space. The noncontinuum processes reflect, among other things, deceleration of vehicles due to interactions between vehicles from the same and from different user classes, immediate lane changing, postponed lane changing by constrained drivers, and spontaneous lane changing by free-flowing drivers. Finally, it is shown how the multilane-multiclass model relates to traditional aggregate-lane and mixed-class gas–kinetic equations. This is accomplished by aggregation across the motorway lanes and the user classes. It is shown that the expression reflecting vehicle interactions in traditional models is only valid for dilute traffic. A remedial expression is proposed.

Recent and some new results of observations of traffic flow dynamics in congested conditions are considered. On the basis of these results, hypotheses about properties of congested flow are proposed and discussed. In particular, it is proposed that the complexity of congested flow is linked to (a) the existence of three qualitatively different phases in traffic flow: free flow, synchronized flow, and wide traffic jams; (b) an occurrence of three qualitatively different types of phase transitions between these three phases of traffic flow; (c) the result that all these qualitatively different phase transitions are related to the same class of “first-order local phase transitions;” (d) complex dynamical behavior of synchronized flow; and (e) the existence of two qualitatively different kinds of “nucleation effects” in traffic flow: the nucleation effect responsible for the jam’s formation, which is linked to an avalanche increase in the amplitude of an initial local perturbation of traffic variables (vehicle speed, density, or both), and the one responsible for the phase transition from free flow to synchronized flow, which is linked to an avalanche decrease in the probability of passing in free traffic flow.

In previous real-time flow prediction studies, the emphasis was placed on the prediction accuracy of the model. The accuracy of the prediction bounds (or limits), on the other hand, was largely ignored. Prediction bounds are, however, important input parameters in such applications as real-time stochastic traffic control, incident detection, and route guidance in the context of dynamic traffic assignment. The objectives of this study are to explore the statistical nature of traffic flows when aggregated at short time intervals and to examine the potential of using the generalized linear model in the dynamic setting to predict traffic flows and provide prediction bounds. Specifically, this study derives recursive algorithms based on the quasi-likelihood principle and performs on-line, multiple-step-ahead predictions of short-term arrival flows for signalized intersections. Preliminary results are presented using a simulated data set from CORSIM and a real data set collected from signalized intersections.

Traffic volume is one of the fundamental types of data that have been used for the traffic control and planning process. Forecasting needs and efforts for various applications will be increased with the deployment of advanced traffic management systems. With the importance of the short-term traffic forecasting task, numerous techniques have been utilized to improve its accuracy. The use of the subset autoregressive integrated moving average (ARIMA) model for short-term traffic volume forecasting is investigated. A typical time-series modeling procedure was employed for this study. Model identification was carried out with Akaike’s information criterion. The conditional maximum likelihood method was used for the parameter estimation process. Two white noise tests were applied for model verification. From the analysis results, four time-series models in different categories were identified and used for the one-step-ahead forecasting task. The performance of each model was evaluated using two statistical error estimates. Results showed that all time-series models performed well with reasonable accuracy. However, it was observed that the subset ARIMA model gave more stable and accurate results than other time-series models, especially a full ARIMA model. It is believed that the use of a subset ARIMA model could increase the accuracy of the short-term forecasting task within time-series models.

Cowan’s M3 distribution has been used in several studies on unsignalized intersections, especially roundabouts. It allows separate analysis of follower and nonfollower headways, and it is simple enough for models of considerable complexity. Although M3 gives a simplistic model for short headways, it has many desirable properties as a headway distribution. It is possible to estimate its parameters with simple methods and with reasonable accuracy, and the estimated models give good results in the capacity analysis of unsignalized intersections. Described in this paper are the basic statistical properties of M3, and the accuracy of the method of moments and least-squares parameter estimators is evaluated. The validity of M3 in unsignalized intersection capacity analysis is evaluated by a Monte Carlo simulation with samples from M3 and semi-Poisson distributions.

Simulation of a large integrated (street and freeway) network with three state-of-the-art software programs is presented. The 20-centerline-km network includes three on-ramps, three off-ramps, an on-and off-ramp weaving section, and a high design arterial with 11 signalized intersections. After considerable modifications to default settings, all three software programs were able to replicate field-measured volumes well. INTEGRATION required extensive modifications to approximate complex signal timing plans and had problems with lane alignment on the wide arterial. CORSIM’s FRESIM component had difficulty with vehicles that miss their destination and required car-following parameter settings corresponding to unusually high capacities to produce good results. WATSim needed the fewest modifications, and it was primarily sensitive to merging and acceleration lengths. WATSim and CORSIM speeds were close to each other. INTEGRATION’s simplified treatment of signalization produced higher street link speeds.

Recent studies have identified the median U-turn (MUT) design as having potential for improving intersection stopped delay at isolated intersections. This study was an effort to quantify the reductions in travel time and other critical traffic operation measures of effectiveness for the MUT and super-street median (SSM) geometric designs over a system of signals compared with the traditional two-way left-turn lane (TWLTL) design. The key function of the MUT design is removal of all left-turn movements at signalized intersections, creating two-phase signal operations and increased progression opportunities. The SSM design allows perfect progression of through traffic in both directions because signals on both sides of the arterial can operate independently. The analysts constructed models of a typical suburban arterial corridor near Detroit, Michigan, in CORSIM according to a fixed external-node coordinate system and fixed origin-destination volumes. These allowed an equitable systemwide comparison of measures of effectiveness between the arterial designs. An analysis of variance determined the importance of arterial geometry related to total system time, average stops per vehicle, and average speed. Results showed that the MUT and SSM designs improved system travel time and average speed compared with the TWLTL design. Analyzing geometry as a function of the time of day showed that the MUT and SSM alternatives significantly reduced system travel time and increased average speeds during the a.m. and p.m. peak hours and showed very similar results compared with the TWLTL design during off-peak conditions. The authors recommend that engineers analyzing alternatives for arterials similar to that tested consider the MUT and SSM designs.

The effects of architecture, learning mode, and learning rate on the performance of a level-of-service (LOS) analysis model using an artificial neural network (ANN) are discussed. Multilayer LOSANN models demonstrated improved quality of learning and testing over single-layered models in evaluating level of service of signalized intersections given geometric, traffic, and traffic signal control data. At present, LOSANN takes delay data from Highway Capacity Software (HCS) outputs; hence its accuracy is constrained by the accuracy of the HCS analyses. However, if delays can be determined directly by field observation, the relationships (or patterns) between field-measured delays and the traffic, geometric, and signal control conditions can be fed to LOSANN. Then the neural network-based model can evaluate the level of service at a higher level of accuracy, and such models can be used as part of advanced traffic management systems to automate LOS analyses.

The arrival times of vehicles traveling southbound along a two-lane, bidirectional highway were recorded at eight neighboring locations upstream of a bottleneck caused by an oversaturated traffic signal. Cumulative curves constructed from these observations describe completely and in great detail the evolution of the resulting long queues. These queues formed directly upstream of the signal when the signal’s service rate fell below the southbound arrival rates, and never formed away from the bottleneck. The predictability of bottlenecks like the one studied here can be exploited to manage traffic more effectively. The behavior of vehicles within the queue, however, was rather interesting. Although the flow oscillations generated by the traffic signal were damped out within 0.8 km (0.5 mi) of the bottleneck, it was found that other oscillations arose within the queue farther upstream, at varied locations, and then grew in amplitude as they propagated in the upstream direction. Thus, the queue appeared to be stable close to the bottleneck and unstable far away. Oscillations never propagated beyond the upstream end of the queue, however; that is, the unusual phenomena always arose after the onset of queueing and remained confined within the queue. Some of these findings run contrary to current theories of traffic flow. Because the data set collected in this study is unprecedented in scope and detail and so that it may be of use to other researchers, it has been posted on the Internet and is fully described here.

The composition of traffic in developing countries is mixed, with a variety of vehicles, motorized and nonmotorized, using the same right-of-way. The motorized or fast-moving vehicles include passenger cars, buses, trucks, auto-rickshaws, scooters, and motorcycles; nonmotorized or slow-moving vehicles include bicycles, cycle-rickshaws, and animal-drawn carts. The proportion of slow-moving vehicles in the traffic stream may vary from 10 to 80 percent. Since the 1950s, considerable research has been done to develop traffic flow models for roadways with mainly motorized vehicles or homogenous traffic, representing the composition of traffic primarily in industrialized countries. A review of the literature has shown that limited studies have been done to develop an understanding of traffic flow for non-lane-based heterogeneous or mixed traffic conditions in developing countries. Some efforts have applied a variation of practices developed for homogenous traffic by converting heterogeneous traffic to equivalent passenger-car units and then applying procedures for homogenous traffic. However, these efforts have produced mixed results. A comprehensive review is presented of the few studies on mixed, non-lane-based traffic in India, Bangladesh, and Indonesia. The unique characteristics of traffic composition, driver behavior, roadway geometry, maneuverability, and vehicular interactions are also presented. Most of the studies conducted in this area are not often readily available. This paper, therefore, should serve as a good reference.

This paper is an extension of earlier work on dynamic traffic assignment with an analytically embedded traffic flow simulation. Previously, a modified Greenshields’ speed-density relationship was used to derive a link travel time function that is monotonically increasing and convex with respect to density. It was recognized that this link travel time function is more applicable for freeway traffic than for arterial street traffic, where a large portion of travel times occurs at the nodes as a result of queueing. In this paper a version of the model is developed with explicit queueing links to simulate such traffic conditions on an arterial street network. The dynamic traffic assignment problem is formulated in a two-level optimization framework, separating the temporal and spatial variables into two problems. The lowerlevel problem represents an equilibrium traffic assignment model that solves the spatial variables subject to a set of feasible paths fixed by the temporal estimates from the upper-level problem. The upper-level problem determines the temporal variables by finding the minimum time-dependent travel times from the origin to all other nodes, assuming equilibrium traffic loads assigned from the previous lower-level problem to be fixed. A set of recursive equations is derived to compute the link traversal times that satisfy the first-in, first-out requirement. An iterative solution algorithm using small time steps with queueing links is developed to solve the proposed two-level dynamic traffic assignment problem. Results are also provided to show the effects of embedding a traffic simulation with explicit treatment of queues to capture traffic dynamics on arterial streets.

Six finite-difference equations often used to approximate the continuity equation of traffic flow are compared in terms of their ability to solve the Lighthill-Whitham-Richards model. Ten representative initial traffic flow conditions were selected, and numerical solutions were obtained for the selected cases using the six finite-difference equations. Results are compared with analytical solutions, and the numerical properties of these approximations are examined. It was found that the upwind finite-difference schemes with volume corrections (Godunov-type schemes) produce the best results for all 10 cases and are reasonable finite-difference schemes for simulating traffic flow.

Protected-permitted phasing from shared-use lanes (i.e., no left-turn bays) exists in certain situations where space and other constraints make this measure expedient in the face of growing traffic demand. An alternative method for handling this special case is proposed as an enhancement to the existing treatment in the 1997 update to the Highway Capacity Manual. The method proposed is logical and consistent. The more practical issue of whether it is also more accurate is addressed in a series of simulation cases. Simulation was used to provide the reference conditions because of field costs, number of variables to be controlled, and the fact that a well-established model exists. The results show that the existing method systematically overestimates lane group capacity by 10 to 25 percent for this special case, whereas the proposed alternative estimates the lane group capacity by ±10 percent or less.

Unsignalized intersections are the most common intersection type. They can have high cost implications for the community. A new development often has an unsignalized intersection for ingress and egress, and the basis for deciding whether an upgraded intersection is needed is based on the performance of that intersection. The concept of limited priority, in which the major-stream drivers slow and allow minor-stream drivers to enter in front of them if the average delay to the major-stream vehicles is likely to be short, is addressed. Similar behavior could also be caused by minor-stream drivers forcing their way into a priority stream. This behavior leads to a more efficient performance of the merge. Relationships to predict the capacity and the maximum delay to major-stream vehicles when drivers exhibit this limited-priority behavior are developed. The implications of the capacity of this merge type are evaluated.

Automated incident detection algorithms usually utilize direct measurements of traffic volumes and occupancies as input and output binary indications of the traffic status for every time step. The transferability of such algorithms is often poor because of the extensive overlap between the incident-related cluster and the incident-free cluster of patterns in the input space. The observed overlap is not only significant but also location specific. The extent of the overlap is shown, its causes are discussed, and an input preprocessor is proposed that relies on modifying the input feature space in order to segregate the two clusters of patterns and reduce their overlap. The result is a more readily separable set of classes; easier classification of new patterns, which leads to better performance; and most important, less site-dependent performance and hence significantly better direct transferability of the detection algorithm with less need for retraining at new sites. On the output side, isolated binary algorithm outputs also have several drawbacks that are identified. An output postprocessor is proposed that links the isolated outputs using a Bayesian update process and converts the isolated outputs into a continuous probabilistic measure that is updated for every time step. The output at one interval is therefore used as a prior probability for the next. The preprocessor and the postprocessor proposed in this paper were developed as integrated components of a universally transferable freeway incident detection framework, UNITID, but are generally applicable to any similar incident detection system.