A Vehicle-Group-Based Traffic Control Approach at Isolated Signalized Intersections in Mixed Traffic Environments

Over the past few decades, traffic signals have played a crucial role in urban intersection control. With the advent of connected and autonomous vehicles (CAVs), traffic control at intersections can be further enhanced through the precise planning of CAV trajectories. In this article, we propose a vehicle-group-based traffic control approach to optimize both signal timing and vehicle trajectories in mixed traffic environments with both CAVs and human-driven vehicles (HDVs). This approach builds upon centralized traffic control approaches, while incorporating the merits of decentralized vehicle control methods. A two-stage optimization framework is proposed. In the first stage, signal timing is optimized considering all vehicles in the control area. To address computational issues, a dynamic programming algorithm with a greedy assignment strategy is designed. In the second stage, vehicles are organized into vehicle groups with CAVs leading HDVs. The CAV trajectories are optimized at the group level. The concept of vehicle group does not require strict leader-follower relationships or designated cruising states, thus enhancing the flexibility in implementation at intersections. To account for the stochastic nature of HDV behaviors, an implementation framework for both traffic signal and vehicles is designed, incorporating an asynchronous receding horizon control scheme and multiple execution modes. Comprehensive comparative experiments have been conducted involving the proposed method and eight different baselines, across a wide range of scenarios, including varying traffic demands, different market penetration rates (MPRs) of CAVs, as well as various levels of randomness. The experimental results not only show that the proposed method outperforms existing benchmarks in terms of vehicle delay and fuel consumption, but also demonstrate that it can achieve coordination among vehicles, leading to improved performance compared to decentralized vehicle control, while enhancing computational efficiency in comparison with centralized control methods.