Journal of Transportation Engineering, Part A: Systems

Accurately evaluating the safety effects of autonomous vehicles (AVs) has become more pressing with the increased adoption rate of AVs. This study utilizes a multiagent adversarial inverse reinforcement learning (MAAIRL) framework for modeling the interactions between AVs and pedestrians in four different cities: Boston, Las Vegas, Pittsburgh, and Singapore. Multiagent actor-critic with Kronecker factors deep reinforcement learning (MACK DRL), a paradigm that extends deep reinforcement learning (DRL), was used to model the behavior of both AVs and pedestrians and to determine their policies and collision avoidance strategies. Simulated trajectories are compared to actual trajectories and the results are evaluated to analyze the behavior of both AVs and pedestrians in terms of their evasive actions such as swerving, accelerating, or decelerating. The multiagent model provides a more comprehensive insight into how road users act in situations of conflict and accounts for changes in the environment. The study also shows that the level of competition between AVs and pedestrians varies significantly across different cities. Las Vegas has the most competitive relationship between AVs and pedestrians, while Singapore has the least competitive environment. The study also highlights the importance of cooperative behavior, particularly in yielding to pedestrians, in reducing the level of competition between AVs and pedestrians. In summary, this research provides valuable insights into the behavior of AVs and pedestrians and can be used to inform the development of more efficient and safe autonomous mobility systems.

Under the condition of urban rail transit uncertainty of passenger demand and the high frequency of departure intervals, this study presents an innovative real-time urban rail transit (URT) train service scheduling control framework. In the context of a bidirectional urban rail transit line, a high-fidelity urban rail transit simulation environment was constructed. Within this environment, an advantage actor–critic (A2C) reinforcement learning approach was utilized to train a suitable strategy aimed at minimizing both passenger waiting costs and transit authority operational expenses. Subject to specific constraints, the strategy is designed to generate real-time train schedule based on the representation of traffic state using station congestion levels and train positions. Experimental results on Lines 3 and S7 of Nanjing Metro demonstrated the agent’s effectiveness in achieving high-performance schedules across various scenarios. This research integrates deep reinforcement learning into the optimization of dynamic traffic systems, showing great potential for enhancing the efficiency and resilience of urban transport systems.

Traffic congestion is a serious problem in the US, and traffic signal control is one of the effective solutions to congestion. Previous research on model predictive control (MPC)-based traffic signal control showed substantial benefits over conventional methods. This study focused on implementing MPC over a large-scale network with complex intersections and the impact of cycle length, network size, and imperfect state estimation on performances. This study implemented a virtual phase link (VPL)-based model predictive control method which used the number of vehicles in each VPL as input state variables and was suitable for National Electrical Manufacturing Association (NEMA) ring-barrier control. To test the impact of network size, the performance of distributed MPC (36 intersections in the network are divided into five subnetworks) was compared with that of MPC over the full network for a set of cycle lengths. To test the impact of imperfect state estimation, we synthetically infused estimation error and developed two scenarios, MPC-error and MPC-error narrow, which had higher and lower estimation errors, respectively. The performance of these MPC methods was compared with that of the existing time-of-day (TOD) method and an offline method that used Webster’s method for split and MULTIBAND for cycle length and offset optimization. Trajectory and linkwise signal performance measures were collected from the simulation to evaluate performance. The distributed MPC method with perfect state estimation had the lowest delay and highest energy efficiency of all the methods. The performance of MPC decreased as the prediction inaccuracy increased. MPC-error had 7% and 11% more delay than MPC-error narrow in the morning and evening peaks, respectively. Overall, simulation results suggest that even with imperfect state estimation, MPC methods will outperform offline methods significantly.

The aim of this work is to investigate dynamic division of highway control subareas based on improved Label Propagation Algorithm (LPA) for implementing different signal control strategies in the subareas to improve operation efficiency. A highway network is used as the studied object, where each road segment is defined as a node and the network is divided into three types of nodes: only mainline segment, an on-off-ramp and its connecting mainline, and an overpass. The system is abstracted as a topological set of nodes and their relationships. The concept of multiroad segment correlation degree with predictive function of congestion and dissipation ranges is proposed. The correlation degree model of multiroad segments based on Pearson Correlation Coefficient (PCC) and the importance degree model of road segment based on technique for order preference by similarity to ideal solution (TOPSIS) are established, respectively. Then a dynamic division method of control subareas for highway network is proposed based on improved LPA, where the label updating model based on the importance degree of road segment is established by semisupervised classification algorithm. Multiple scenarios were discussed using the highway network of Baoding City in Hebei Province of China as an example. The results show that the number of subareas is moderate, the internal state is uniform, and the calculation is simple and efficient. The control effect is remarkable especially when the vehicle distribution is uneven. Total delay time is reduced by 10.47%, 4.38%, 4.24%, 9.73%, and 2.33%, respectively, and total travel time is reduced by 12.20%, 7.15%, 7.02%, 11.85%, and 3.78%, respectively, compared with no control method, global control method, the static division method by clustering algorithm, the method by normal LPA algorithm, and the method by Newman algorithm.

Abstract

Macroscopic fundamental diagram (MFD)-based perimeter control has a potential to improve traffic throughput and relieve the overall congestion. In practice, not only the MFD but also the traffic demand suffers from a variety of inherent uncertainties. Hence, this paper aims to contrive a robust control technique to address the perimeter traffic flow control issue in the oversaturated region. To this end, an integral sliding mode control is proposed to drive the traffic state to the desirable condition. The developed strategy deals with practical issues such as finite time stability and range of uncertainty in both MFD and traffic demand. Theoretical analysis verifies that the developed integral sliding mode control approach can guarantee the finite time convergence of the traffic state to the desired one. The performance of the developed scheme is attested by considering various traffic scenarios with uncertain MFDs and traffic demand, in which a comparative performance study with the proportional and integral control method is conducted. It is indicated that the developed scheme can alleviate the congestion in the urban network and improve the throughput of the urban network.

Practical Applications

The research presented in this article proposes a novel perimeter control technique designed to enhance traffic flow management in densely populated urban road networks, particularly during peak congestion times. By leveraging the macroscopic fundamental diagram (MFD) and integral sliding mode control theory, the method significantly boosts traffic throughput and alleviates the impact of congestion. This advancement can be seamlessly integrated into current urban traffic management systems, leading to more efficient and convenient urban commutes. Additionally, it addresses the rise in vehicle emissions due to heavy traffic, aiding in the reduction of pollution and supporting sustainable development initiatives. The technique also curtails vehicle fuel consumption and cuts down on costs associated with congestion-related delays, thereby fostering resource optimization and economic growth within urban environments. Moreover, the method is designed to be resilient, taking into account the variability of traffic demand and ensuring the robustness of the traffic control strategy. Overall, the article demonstrates the practical application of advanced control technologies to address real-world traffic challenges, providing a scalable and robust solution adaptable to realistic situations with the uncertainties of urban traffic demand.

Reliable crash data are not usually available in low- and middle-income countries (LMICs). In these regions, surrogate safety measures (SSMs) can be used as effective tools for quantifying road safety. Traffic conflicts are the most-used SSMs, primarily defined based on temporal or spatial proximity between vehicles. Time-to-collision (TTC), a conflict indicator and its derivatives, are commonly used to define conflicts based on 1D interactions may not be suitable for nonlane-based traffic where vehicular interactions are 2D (longitudinal and lateral). This study aims to propose a methodology to define conflicts considering 2D vehicle interactions. Traffic video data were recorded at four unsignalized T-intersections, identified as black spots on divided highways in India. A bivariate extreme value approach was used to define conflict in 2D vehicular interaction using TTC and lateral gap. The results show that incorporating lateral and longitudinal conflict indicators into the bivariate extreme value models can significantly improve conflict-based risk assessment. The proposed approach can be used to define safety-critical events required in vehicle warning systems for nonlane-based traffic.

Policymakers demonstrate a keen interest in understanding risky driving behaviors to formulate effective countermeasures aimed at reducing accidents and economic losses. With the increasing deployment of millimeter-wave (MMW) radars on roadways, there exists a viable opportunity to gather extensive vehicle information at big data levels from individual drivers traversing through the radar detection range. This study endeavors to analyze traffic flow characteristics and identify risky driving behaviors using the noisy raw vehicle position and speed profiles obtained from MMW radars installed on a highway in China. A series of data cleaning procedures are meticulously implemented to address several typical trajectory errors stemming from MMW radars. Subsequently, after data cleaning, the study identifies risky driving behaviors through established methods found in the literature and evaluates the prevalence of these behaviors across different times of day and days of the week. This research mitigates the gap between raw vehicle trajectories from MMW radar and popular existing risk analysis methods. In addition, this research analyzes the temporal pattern of different risks and pinpoints their inherent connections. The outcomes of this research endeavor hold the potential to furnish practical insights for the formulation of targeted safety enhancement policies by governmental bodies or relevant agencies.

Road safety on horizontal curves has become a great concern due to issues related to the number of road crashes where collisions result in twice as many fatalities as on a straight segment of road. This study is an attempt to improve safety at horizontal curves. The analysis methodology consists of two parts: first, using ArcGIS software, horizontal curves with a right of way of 30, 45, and 60 m and clustering of fatal crashes were identified. The next step used a reliability theory to account for how safe the existing curve is and what the probability of failure or noncompliance will be. Analysis of the traffic crashes over the Delhi map shows that 34 curves are at high risk based on the density of crashes at curves and required road geometry parameters. It was found that, as the curve length increases, the reliability index based on stopping sight distance increases. The reliability index based on radius and superelevation increases with the radius of the curve. The increase in curve length has a much higher effect on safety performance than the curve radius in an urban area. This research will not only help in determining the unsafe curves present but also in taking appropriate measures to lower crash rates.