Car-Following (CF), as a fundamental driving behaviour, has significant influences on the safety and efficiency of traffic flow. Investigating how human drivers react differently when following autonomous vs. human-driven vehicles (HV) is thus critical for mixed traffic flow. Research in this field can be expedited with trajectory datasets collected by Autonomous Vehicles (AVs). However, trajectories collected by AVs are noisy and not readily applicable for studying CF behaviour. This paper extracts and enhances two categories of CF data, HV-following-AV (H-A) and HV-following-HV (H-H), from the open Lyft level-5 dataset. First, CF pairs are selected based on specific rules. Next, the quality of raw data is assessed by anomaly analysis. Then, the raw CF data is corrected and enhanced via motion planning, Kalman filtering, and wavelet denoising. As a result, 29k+ H-A and 42k+ H-H car-following segments are obtained, with a total driving distance of 150k+ km. A diversity assessment shows that the processed data cover complete CF regimes for calibrating CF models. This open and ready-to-use dataset provides the opportunity to investigate the CF behaviours of following AVs vs. HVs from real-world data. It can further facilitate studies on exploring the impact of AVs on mixed urban traffic.@en

This paper deals with the derivation of analytical formulae to estimate the effective capacity at freeway merges in a multilane context. It extends two previous papers that are based on the same modeling framework but that are restricted to a single lane on the freeway (or to the analysis of the right lane only). The analytical expression for the one-lane capacity is recursively applied for all lanes. Lane-changing maneuvers (mandatory for the on-ramp vehicles and discretionary for others) are divided into two non-overlapping local merging areas. Discretionary lane-changes are transformed into a lane-changing flow using an appropriate analytical formula. This defines a system of equations whose unknowns are the capacity on all lanes and the inserting flow coming from the on-ramp. A sensitivity analysis shows that vehicle acceleration and the truck ratio are the most influential parameters for the total capacity. The analytical formulae are proven to match with numerical results from a traffic simulator that fully describes vehicle dynamics. Finally, they provide very good estimates when compared to experimental data for an active merge on the M6 freeway in UK.@en

Perimeter traffic flow control based on the macroscopic or network fundamental diagram provides the opportunity of operating an urban traffic network at its capacity. Because perimeter control operates on the basis of restricting inflow via reduced green times at selected entry (gated) links, vehicles on those links may be subject to queuing and delay. The experienced delay or resulting queue lengths depend on the adopted policy for the distribution of the inflows and corresponding green times at the gated links. The chosen policy may have a significant impact on the traffic system under control. For example, managing queue lengths may reduce the interference with upstream traffic whereas the management of delays may improve users’ perception with respect to equity and fairness. In this paper, an approach has been proposed to distribute the gated flow based on the queue lengths or experienced delay at the gated signalized junctions. This is in contrast to standard practice that distributes inflows proportionally to the gated links’ saturation flows. Perimeter control is then evaluated in a microscopic simulator for a realistic traffic network and compared in three configurations against fixed-time: perimeter control without queue or delay management; perimeter control with relative queue balancing; and perimeter control with delay balancing. It has been found that managing the queues at the gated links not only improves the overall network performance but also reduces the possibility of queue propagation to the upstream junctions. This improves traffic flow outside the protected network by managing the queue propagation at the gated links and reducing the possibility of queue spill-back to upstream intersections. In addition, the results indicate that perimeter control with delay balancing has a similar performance as the case without queue or delay management being a suitable approach for flow distribution among the gated links. In the scenarios with perimeter control with either queue or delay balancing the gap between the ordered flow by the controller and the actual flow crossing the stop-line at the gated links reduced remarkably.@en

This paper presents a novel approach to improve traffic throughput near diverge and weave bottlenecks in mixed traffic with human-driven vehicles (HDVs) and connected automated vehicles (CAVs). This is done by the strategic assignment of CAVs across lanes. The main principle is to induce strategic and necessary lane changes (LCs) (by CAVs and HDVs) well upstream of the potential bottleneck, so that the traffic flow approaching the bottleneck is organized and exhibits fewer throughput-reducing LCs at the bottleneck. A hybrid approach is used to investigate the problem: macroscopic analytical approach to formulate lane assignment strategies, and numerical simulations to quantify the improvements in throughput for various scenarios. Several strategies are formulated considering various operational conditions for each bottleneck type. Furthermore, compensatory behaviour of HDVs in response to the flow/density imbalance created by the CAV lane assignment is explicitly accounted for in our framework. Evaluation by numerical simulations demonstrates significant benefits of the proposed method, even at low to moderate CAV penetration rates: they can lead to an increase of throughput by several percent, thereby decreasing delays significantly.@en

Microscopic traffic flow models enable predictions of traffic operations, which allows traffic engineers to assess the efficiency and safety effects of roadway designs. Modeling vehicle trajectories inside intersections is challenging because there is an infinite number of possible paths in a two-dimensional space, and drivers can simultaneously adapt their speeds as well. To date, human driver models for simultaneous longitudinal and lateral vehicle control based on the infrastructure characteristics and interactions with other drivers inside an intersection are still lacking. The contribution of this paper is threefold. First, it proposes an integrated microscopic traffic flow model to describe human-driven vehicle maneuvers under interactions. Drivers plan their heading and acceleration in the predicted future to minimize costs representing undesirable situations. The model works with a joint optimization for an interaction cost term. The weights associated with the interaction cost reflect how selfish or altruistic drivers are. Second, the proposed model endogenously gives the order of vehicles in case of crossing paths. Third, the paper develops a clustered validation method for microscopic traffic flow models with interacting vehicles, which account for interdriver variations. Results show that the model can accurately describe vehicle passing orders of interacting maneuvers, paths, and speeds against empirical data. The model can be applied to assess various intersection designs.@en

Accurate and explainable short-term traffic forecasting is pivotal for making trustworthy decisions in advanced traffic control and guidance systems. Recently, deep learning approach, as a data-driven alternative to traffic flow model-based data assimilation and prediction methods, has become popular in this domain. Many of these deep learning models show promising predictive performance, but inherently suffer from a lack of interpretability. This difficulty largely originates from the inconsistency between the static input–output mappings encoded in deep neural networks and the dynamic nature of traffic phenomena. Under different traffic conditions, such as freely-flowing versus heavily congested traffic, different mappings are needed to predict the propagation of congestion and the resulting speeds over the network more accurately. In this study, we design a novel variant of the graph attention mechanism. The major innovation of this so-called dynamic graph convolution (DGC) module is that local area-wide graph convolutional kernels are dynamically generated from evolving traffic states to capture real-time spatial dependencies. When traffic conditions change, the spatial correlation encoded by DGC module changes as well. Using the DGC, we propose a multistep traffic forecasting model, the Dynamic Graph Convolutional Network (DGCN). Experiments using real freeway data show that the DGCN has a competitive predictive performance compared to other state-of-the-art models. Equally importantly, the prediction process in the DGCN and the trained parameters are indeed explainable. It turns out that the DGCN learns to mimic the upstream–downstream asymmetric information flow of typical road traffic operations. Specifically, there exists a speed-dependent optimal receptive field – which governs what information the DGC kernels assimilate – that is consistent with the back-propagation speed of stop-and-go waves in traffic streams. This implies that the learnt parameters are consistent with traffic flow theory. We believe that this research paves a path to more transparent deep learning models applied for short-term traffic forecasting.@en

For better and more efficient motorway traffic control strategies on a lane level, accurate modelling and prediction of traffic conditions is essential. Existing real-time traffic state estimation methods aggregate traffic across lanes. Hence, there is a need for lane-specific traffic flow models. The majority of the existing macroscopic lane-level traffic flow models use speed difference among lanes to explain lane change decisions. One of the major disadvantages of using speed difference as motivation for lane changing is the different speeds at which the lanes on a motorway operate. These models which are usually extensions of the Cell Transmission Model (CTM) also suffer in the manner in which lateral flow is transferred among cells in different lanes. In order to overcome these limitations, a model is proposed using density difference as an incentive to compute lane change rates rather than speed. A two-step transfer of lateral flow among cell segments is also considered where the lateral demand of a cell is dependent upon the receiving capacity of the adjacent and downstream cells in the target lane. The advantages of the presented approach is shown by comparing the proposed model with a conventional model. It was observed that the proposed model performed better in terms estimating densities across lanes. The proposed model is a step in the development of more accurate lane-specific traffic flow models.@en

Traffic flow downstream of the congestion is generally lower than the pre-queue capacity. This phenomenon is called the capacity drop. Recent empirical observations show a positive relationship between the speed in congestion and the queue discharge rate. Literature indicates that variations in driver behaviors can account for the capacity drop. However, to the best of authors' knowledge, there is no solid understanding of what and how this variation in driver behaviors lead to the capacity drop, especially without lane changing. Hence, this paper fills this gap. We incorporate the empirically observed desired acceleration stochasticity into a car-following model. The extended parsimonious car-following model shows different capacity drop magnitudes in different traffic situations, consistent with empirical observations. All results indicate that the stochasticity of desired accelerations is a significant reason for the capacity drop. The new insights can be used to develop and test new measures in traffic control.@en

A two-Fluid Model (TFM) of urban traffic provides the macroscopic description of traffic state. The TFMs parameters are hard to calibrate, particularly for the dynamic traffic conditions. This leads to the TFM often being used to compare the quality of service through the plot of stopping time versus trip time of the vehicles in the network. Recently, the taxi GPS data has been applied to predict the traffic condition at the network level. Despite the network-wide coverage of the taxi GPS probe data, the penetration rate of taxis in the network traffic is still a vital and challenging issue for traffic estimation purpose. It is necessary to estimate penetration rate of taxis by combining with other data sources. Here, we propose a novel approach to fill two gaps: TFM parameter calibration and the taxis penetration rate. This method stretches the description of TFM to a zone size. The method is applied to real Changsha city GPS data, calibrating the parameters. The macroscopic fundamental diagram of the large-scale city is derived. For the Changsha case, running speed is the super-linear power law of the fraction of running cars; the fraction of stopping time is nearly linear power law of density, which can be an alternative of the density. The proposed method enables the calibration of TFM parameters and macroscopic traffic monitoring at urban scale using only GPS data.@en

Traffic processes cause several problems in the world. Traffic delay, pollution are some of it. They can be solved with the right road design or traffic management (control) measure. Before implementing these designs of measures, though, their effect could be tested. To this end, knowledge of traffic flow theory is needed.@en

Traffic congestion leads to delays and increased carbon dioxide (CO2) emissions. Traffic management measures such as providing information on environmental route costs have been proposed to mitigate congestion. Multi-criteria routing dynamic traffic assignment (MCR-DTA) models are needed to evaluate the effectiveness of such measures. This paper presents a simulation-based bi-level optimization method to solve the MCR-DTA problem, which works as follows. Route costs include travel times and emissions, but those are updated inside two different loops. In the inner loop, emission costs are considered fixed; the assignment is performed by updating route travel times, using a traditional DTA tool. Then, in the outer loop, emissions are calculated based on link loads and fed back to the DTA tool, which performs a new assignment. The MCR user equilibrium is found when emissions or predefined generalized costs converge to an equilibrium. The bi-level method is first tested on a small network, showing that the proposed method is able to effectively solve the MCR-DTA problem. Next, the method is applied to a medium-size urban network. The results show that if drivers choose routes based on emissions besides travel time, the average travel time and emissions per vehicle decrease. This occurs because congested links have a higher impact on route costs; hence the equilibrium is pushed away from the single-criteria routing (SCR) user optimum towards the SCR system optimum. Results support the conclusion that informing drivers about CO2 emissions per route can potentially lead to decreased delay and emissions in real networks.@en

Pedestrians are among the travelers most vulnerable to collisions that are associated with high fatality and injury rates. The increasing rate of urbanization and mixed land-use construction make walking (along with other non-motorized travel) a predominant transportation mode with a wide variety of behaviors expected. Because of the inherent safety concerns seen in pedestrian transportation infrastructures, especially those with conflicting multimodal movements expected (crosswalks, transit platforms, etc.), it is important that pedestrian behavior is modeled as a risk-taking stochastic behavior that may lead to errors and thus collision formation. In previous work, the complexity and cost associated with building pedestrian models in a cognitive-based environment weighted down the construction of simulation tools that can capture pedestrian-involved collisions, including those seen in shared space environments. In this paper, a tool that will help evaluate the safety of pedestrian traffic is initiated: an extended modeling framework of pedestrian walking behavior is adopted while incorporating different physiological, physical, and decision-making elements. The focus is on operational decisions (i.e., path choices defined by longitudinal and lateral trajectories) with a pre-specified set of origins and destinations. The model relies on the prospect theory paradigm where pedestrians evaluate their acceleration and directional alternatives while considering the possibility of colliding with other ‘‘particles.’’ Using a genetic algorithm method, the new model is calibrated using detailed trajectory data. This model can be extended to model the interactions between a variety of different modes that are present in different mixed land-use environments.@en

Literature shows that–under specific conditions–the Macroscopic Fundamental Diagram (MFD) describes a crisp relationship between the average flow (production) and the average density in an entire network. The limiting condition is that traffic conditions must be homogeneous over the whole network. Recent works describe hysteresis effects: systematic deviations from the MFD as a result of loading and unloading. This article proposes a two dimensional generalization of the MFD, the so-called Generalized Macroscopic Fundamental Diagram (GMFD), which relates the average flow to both the average density and the (spatial) inhomogeneity of density. The most important contribution is that we show this is a continuous function, of which the MFD is a projection. Using the GMFD, we can describe the mentioned hysteresis patterns in the MFD. The underlying traffic phenomenon explaining the two dimensional surface described by the GMFD is that congestion concentrates (and subsequently spreads out) around the bottlenecks that oversaturate first. We call this the nucleation effect. Due to this effect, the network flow is not constant for a fixed number of vehicles as predicted by the MFD, but decreases due to local queueing and spill back processes around the congestion “nuclei”. During this build up of congestion, the production hence decreases, which gives the hysteresis effects.@en