Transportation Research Record: Journal of the Transportation Research Board

Continuum theory is used to explain why stop-and-go oscillations in congested freeway traffic change their amplitudes when they encounter the vehicular merging and diverging maneuvers that take place near ramps. The theory describes how oscillations diminish in amplitude when they propagate past a queued (and unmetered) on-ramp and how they grow when they propagate past an off-ramp. The premise is that merging (diverging) flows change in response to freeway oscillations and that these changes in flow dampen (amplify) oscillations. The theory's descriptions are simple and rational; all its inputs and outputs are directly observable; and its predictions are shown to match real data. The theory is tested against real data collected over multiple days from congested merge and diverge sites with videos and inductive loop detectors. For merges, predictions are found to agree with observation to within 10%, and for diverges, to within 12%. The paper thus resolves in a simple way a puzzling traffic feature reported in previous studies.

Analysis of public transport system performance and level of service in urban areas is essential. Dynamic modeling of traffic conditions, passenger demand, and transit operations is important to represent adequately the complexity of and the interactions between these components in modern public transportation systems. This paper presents a transit simulation model designed to support evaluation of operations planning and control, especially in the context of advanced public transportation systems. Unlike most previous efforts in this area, the simulation model is built on a platform of a mesoscopic traffic simulation model, which allows modeling of the operation dynamics of large-scale transit systems, taking into account the main sources of service uncertainty and stochasticity. The capabilities of Mezzo as an evaluation tool of transit operations are demonstrated with an application to a real-world, high-demand bus line in metropolitan Tel Aviv, Israel, under various scenarios. The application shows that important phenomena such as bus bunching are reproduced realistically. A comparison of simulated running times and headway distributions with field data shows the model is capable of replicating observed data.

The effect of rain on freeway flow breakdown behavior is investigated. Three aspects of flow breakdown are analyzed for rain versus no rain (clear) weather conditions. First, the probability of breakdown occurrence is examined by analyzing the distribution of prebreakdown flow rates observed immediately before the onset of traffic breakdown by using a survival analysis approach. At all study sections, a reduction with pre-breakdown flow rates is observed under rain conditions compared with distributions under no rain and confirms higher breakdown likelihoods at lower flows. Log likelihood ratio tests confirm the statistical significance of differences in the prebreakdown flow rate distribution parameters under rain compared with clear conditions. Second, breakdown duration is examined by estimating a semiparametric Cox proportional hazard model. With a rain event indicator set as an independent variable, the effect of rain on breakdown duration is observed. Rain during a breakdown episode is found to increase its duration, whereas rain before breakdown does not appear to affect duration. Finally, prebreakdown and postbreakdown flow rates are compared. Overall, while a reduction in prebreakdown flow rates is observed because of rain, the flow drop between prebreakdown and postbreakdown is not much different between rain (3.9% to 12.0%) and no rain (7.8% to 12.7%) conditions.

The kinematic wave model is often used in simulation tools to describe dynamic traffic flow and to estimate and predict traffic states. Discretization of the model is generally based on Eulerian coordinates, which are fixed in space. However, the Lagrangian coordinate system, in which the coordinates move with the velocity of the vehicles, results in more accurate solutions. Furthermore, if the model includes multiple user classes, it describes real traffic more accurately. Such a multiclass model, in contrast to a mixed-class model, treats different types of vehicles (e.g., passenger cars and trucks or vehicles with different origins or destinations, or both) differently. The Lagrangian coordinate system is combined with a multiclass model, and a Lagrangian formulation of the kinematic wave model for multiple user classes is proposed. It is shown that the advantages of the Lagrangian formulation also apply for the multiclass model. Simulations based on the Lagrangian formulation result in more accurate solutions than simulations based on the Eulerian formulation.

Microscopic simulation models have become widely applied tools in traffic engineering. Nevertheless, parameter identification of these models remains a difficult task. This is partially because parameters are generally not directly observable from common traffic data; also there is a lack of reliable statistical estimation techniques. This study puts forward a new general and structured approach to identifying parameters of car-following models. One of the main contributions of this study is joint estimation of parameters for multiple vehicles. Furthermore, prior information on the parameter values (or the valid range of values) can be estimated. The study also deals with serial correlation in the trajectory data. In doing so, the newly developed approach generalizes the maximum likelihood estimation approach proposed by the authors. The approach allows for statistical analysis of the model estimates, including the standard error of the parameter estimates and the correlation of the estimates. With the likelihood ratio test, models of different complexity (defined by the number of model parameters) can be cross-compared. A useful property of this test is that it takes into account the number of parameters of a model as well as the performance. The approach is applied to car-following behavior by using Dutch freeway vehicle trajectories collected from a helicopter.

Travel time unreliability is an important characterization of transportation systems. The appropriate modeling and reporting of travel time reliability is crucial for individual travelers as well as management agencies. A novel multistate model is proposed for travel time modeling and reporting. The model advances travel time modeling in two aspects. First, the multistate model provides much improved model fitting as compared with single-mode models. Second and more important, the proposed model provides a connection between travel time distributions and the underlying travel time state. This connection allows for the quantitative evaluation of the probability of each travel time state as well as the uncertainty associated with each state, for example, the probability of encountering congestion at a given time of day and the expected travel time if congestion is experienced. A simulation study was conducted to demonstrate the performance of the model. The proposed model was applied to field data collected at San Antonio, Texas. The variation of the model parameters as a function of time of day was also investigated. The simulation study and field data analysis confirmed the superiority of the multistate model over the state-of-practice single-mode travel time reliability models and for ease of interpretation.

On a multilane roadway, systematic lane changes can seriously disrupt traffic flow and cause capacity reductions. Therefore it is important to understand the characteristics of lane-changing traffic flow. In a previous study, a new variable, lane-changing intensity, was theoretically derived by doubling the contribution to the total density of a lane-changing vehicle during its lane-changing period, and an intensity–density relation was incorporated into the fundamental diagram to capture such bottleneck effects at the aggregate level. In this study, a new interpretation of lane-changing intensity is proposed by following Edie's definition of traffic density. Detailed studies of macroscopic characteristics of lane-changing traffic with 75 min of vehicle trajectories collected for a freeway section on I-80 near San Francisco, California, by FHWA's next-generation simulation project are presented. About 10% misindentified lane changes are excluded. For different time intervals, cell sizes, and locations, the well-definedness of the lane-changing intensity variable and the intensity–density relation are verified.

The values on parameters describing longitudinal driving behavior in car-following models differ substantially between drivers. Different individual interactions with the environment are assumed to play an important role, which might be explained through mental workload. Therefore a driving simulator experiment with a repeated measures design was performed to investigate to what extent perception of an incident in the other driving lane influences physiological indicators as well as subjective estimates of mental workload and longitudinal driving behavior. As almost none of the current models of car-following behavior incorporate mental workload as a determinant of driving behavior, an investigation was conducted by using a calibration approach for joint estimation to determine whether these models, represented by the intelligent driver model and the Helly model, adequately described longitudinal driving behavior in case of incidents in the other driving lane. The results indicated that perception of an incident in the other driving lane influenced mental workload as measured by physiological indicators and longitudinal driving behavior. In addition, the results indicated that current car-following models did not adequately describe driving behavior in case of incidents in the other driving lane.

Travel time reliability has received much attention from researchers and practitioners because it not only affects driver route choice behavior but also is used in the assessment of transportation system performance. Current state-of-the-art travel time reliability research assumes that roadway travel times follow a unimodal distribution. Specifically, a Weibull, exponential, lognormal, or normal distribution is used within the context of travel time reliability modeling. However, field observations demonstrate that roadway travel times are multimodal, especially during peak periods. This paper demonstrates that the multimodes observed in field data are a result of temporal variations in travel times both between and within days. This study evaluates the potential for generating multimodal travel times by using INTEGRATION software and investigates the underlying traffic states that result in these multimodes. The study validates the two-state model proposed in an earlier study and demonstrates the robustness of model parameter estimates under varying stochastic traffic congestion levels and sampling techniques.

This paper presents a methodology for using vehicle trajectory data to study the intradriver heterogeneity of driving behavior between the acceleration process and the deceleration process. Trajectory data were collected during peak hours on Dutch Motorway A2. Criteria were proposed for the selection of subtrajectories corresponding to both the acceleration and the deceleration processes of car-following. With the application of these subtrajectories to calibrate three types of models (the Helly model, the Gipps model, and the intelligent driver model), it was found that obvious intradriver heterogeneities existed in driving behaviors between the acceleration and deceleration processes of car-following: (a) the average response time of drivers in the acceleration process was longer than that in the deceleration process according to the prediction of the Helly and intelligent driver models; (b) drivers were apt to respond more intensively to surrounding traffic in the deceleration process than they did in the acceleration process; and (c) more than 65% of drivers involved in this study drove in obviously different styles between the acceleration and deceleration processes. Moreover, a compensation for the response delay from model parameters was observed, and all three models presented low robustness in predicting driving behaviors of one car-following process with the parameters optimized from the data of other car-following processes. This work not only provides insight into intradriver heterogeneity in car-following behaviors but also suggests some important criteria for car-following modeling.

Low-speed merging maneuvers performed within a free-flow stream are believed to trigger congestion. These accelerating moving bottlenecks introduce local constraints that can disturb the flow at a local or global scale. Low-speed merging maneuvers are also suspected to cause capacity drop. Using the kinematic wave theory, this paper explores the analytical solution of a simple first-order model when moving boundary conditions are introduced. The paper shows that shock waves initiated by low-speed merging maneuvers are a linear transformation of the moving boundary conditions, no matter the shape of the moving boundary condition. These results are then applied to typical situations to show that the interaction of two moving boundaries can modify the analytical solution of the problem. The results are then extended to multiple merging maneuvers to show that they can interact. Every possible interaction between two identical merging maneuvers is explored to identify the conditions that lead to global congestion. Finally, these results are used to propose an analytical formulation of the capacity drop for multiple merging maneuvers at a single location. It is shown that capacity drop is related to the demands on the minor and major streams and to the speed of the merging vehicle.

The ability of microscopic (simulation) models to represent lane-changing behavior according to reality has recently been questioned. In this paper the merging maneuver (a specific type of lane changing) is analyzed with empirical data. First, a conceptual model is composed; it includes the factors influencing merging behavior, namely the merge location and its relation to prevailing driving conditions, gap acceptance, and the relaxation phenomenon. The empirical data set consists of 35 min of vehicle maneuvers on 400 m of freeway, collected by a camera mounted underneath a helicopter. This process results in a data set of 3,459 vehicle trajectories, from which 704 trajectories describe merging vehicles. It is found that different merge locations are used under congested and freeflow traffic conditions. During free-flow, most vehicles merge at the first half of the acceleration lane. Under congested traffic conditions, relatively more merges are registered at the end of the acceleration lane. The smallest accepted gap observed in the data set lies between 0.75 and 1.0 s. Net headways between the merging vehicle and the new leader and new follower of less than 0.25 s are recorded. These short accepted gaps are growing over time and indicate relaxation behavior. From the data analysis it can be concluded that gap acceptance theories, as they are used in current models and theories to model merge behavior, are not able to model the observed behavior accurately.

Traffic state estimation problems must deal with imperfect information. Imperfect information may occur as a result of misalignments between reality and the assumptions or simplifications made when developing the models as well as incomplete or corrupted data. One of the well-known methods of handling imperfect information is the Kalman filter algorithm and its variations, such as the extended Kalman filter and the unscented Kalman filter. While the standard Kalman filter and the extended Kalman filter have been widely applied to the state estimation of linear systems, the unscented Kalman filter has been reported the better choice in the multiclass traffic state estimation. In many applications of Kalman filters to traffic estimation problems, the model and measurement noise covariance matrices are normally estimated. When there is a mismatch between the true and the assumed noise distribution, however, the filter often suffers from performance degradation and even divergence in certain situations. To this end, this paper presents a more efficient and accurate algorithm embedded in the unscented Kalman filter to simultaneously estimate the traffic state and the model noise distribution statistics. The proposed method is facilitated through a simultaneous update of the model noise covariance matrix in the predicted covariance equations of the unscented Kalman filter algorithm. It is found through simulation that the proposed algorithm may improve the model performance over the standard algorithm.

This paper proposes a cellular automata (CA) model incorporating work zone configuration to model work zone traffic. The randomization probability parameter of the proposed CA model is able to characterize driver acceleration–deceleration behavior. The randomization probability should be a function of traffic flow and work zone configuration that comprises the activity length and transition length; however, past studies have assigned a hypothetical constant value for the randomization probability. This paper calibrates the randomization probability from field data, which is determined by minimizing the square error between simulated travel time and observed travel time by using a trial-and-error method. A polynomial regression method is employed to formulate the randomization probability functions in and outside the work zone. A case study was performed to test the proposed CA model dependent on work zone configuration. Comparison of field data and the proposed CA model for travel time and traffic delay shows very close agreement. Statistical comparison of the simulated results from the proposed CA model and PARAMICS indicates that the proposed CA model performs well in modeling work zone traffic.

In developing countries such as India, the traffic on roads is highly heterogeneous in nature, with vehicles of widely varying static and dynamic characteristics. This type of traffic is characterized by lack of lane discipline and free movement over the entire width of the roadway (laneless movement) based on availability of space. In such a regime of traffic flow, the phenomenon of merging of vehicles at junctions of two roads is complex and warrants further study. Merging maneuvers at T-junctions under congested traffic conditions were studied microscopically through video recordings. On the basis of these observations, models for normal and forced merging have been developed. Video data of merging maneuvers were employed to calibrate and validate the models. Such models can be used to simulate highly congested traffic flow in a realistic manner under heterogeneous traffic conditions. They can also give insight on devising better traffic control measures at such junctions.

The stochastic nature of freeway bottleneck breakdown and queue discharge is investigated through a comprehensive analysis of sensor data collected at bottleneck sites in the San Francisco Bay Area, California, and San Antonio, Texas. A new procedure was proposed to define the stochastic variation of the onset of freeway breakdown and of queue discharge capacity on the basis of time-indexed field data of speed–flow profiles. The former was developed as a function of average vehicle time headways preceding observed conditions when both speed was below and density was above locally defined congested flow thresholds. A full-year 15-min data series was used in the demonstration and testing of the procedure and yielded a high degree of statistical confidence in the resulting estimates of headway distribution parameters. The statistical analysis indicated that the probability function of freeway bottleneck prebreakdown headways followed a shifted lognormal distribution. In addition, a recursive queue discharge model was proposed for bottleneck flows under congested (queued) conditions. The proposed queue discharge model was a simple autocorrelated time series recursion that was seeded with the corresponding prebreakdown flow and dampens to the mean queue discharge rate. The proposed stochastic models are robust and accurate and represent a significant improvement in the understanding and modeling of freeway bottleneck flow. The models were implemented in the mesoscopic network model DYNASMART-P to test the effects of stochastic freeway capacity on sustained service rates and network performance.

Regulation of the percentage of visible light transmission (VLT) has been controversial because the driving behavior of the following vehicle depends on the integrity of the vision field. It is conjectured that the level of VLT of a vehicle can affect the stability and headway of vehicles. Car-following theory can be a useful tool to analyze the effect of vehicle VLT on traffic flow in cases in which proper field data can be obtained. Most car-following experiments have been based mainly on mathematical analyses, which, because of a lack of accurate equipment, assumed that the response of the drivers was uniform. Now, global positioning systems (GPSs) and other advanced technologies make it possible to obtain vehicle trajectories and to monitor driving behavior accurately. This study focuses on identifying the effect of vehicle VLT on the local and asymptotic stability underlying car-following theory. Car-following experiments were performed for VLT of 70%, 30%, and 0% with a real-time kinematic GPS receiving apparatus. In addition, both local and asymptotic stability were analyzed to verify the relationship between VLT and the stability of vehicle trajectories. Driving behavior was also investigated with acceleration noise and distance headway. The VLT of tinted windows is found to have a statistically significant effect on the following driver's behavior. Further studies should focus on measuring the negative impact of lower VLT to establish minimum standards for VLT for various velocities and different types of vehicles by using this experimental method.

Estimation of densities on freeways and arterials is critical to traffic control and management. Most previous work, however, focused on freeway density estimation based merely on detector data. This study attempts to estimate traffic density along a signalized arterial by using data from both detectors and a global positioning system (GPS). The approximation to the previously developed MARCOM (Markov compartment) model is adapted to describe arterial traffic states. A hybrid extended Kalman filter is then implemented to integrate the approximated MARCOM with detector and GPS measurements. The proposed model is tested on a single signal link simulated by using VisSim. Test results show that the hybrid extended Kalman filter with GPS data can significantly improve density estimation.

Experimentation and analysis methodologies for data intended for the development of microscopic car-following models are presented. The data methodology centers on the notion that it is better to sample a varied group of real drivers when the subjects are unaware that an experiment is being conducted. This contrasts with the methods used to calibrate most conventional models. The analysis focuses on determining possible causal factors, beyond the kinematic variables most often used, that could play a role in real traffic. These factors include various human characteristics (e.g., gender and distractions related to in-vehicle conditions like telephoning and vehicle occupancy); traffic and road characteristics (e.g., type of vehicle, congestion level, and location of driving lane); and environmental characteristics (e.g., time of day and weather conditions). The paper shows the magnitudes of the relationships that are discovered and offers suggestions as to how these results might contribute to the development of better models.