Joint optimization of vehicle platoon and traffic signal with mixed traffic flow at intersections: Deep reinforcement learning approach

Vehicle formation and the management of right-of-way conflicts at intersections can realize both the one-dimensional and two-dimensional benefits of connected and automated vehicles (CAVs). Although researchers have proposed joint optimization of vehicle trajectories and traffic signals in fully connected environments to exploit these benefits, the advantages of CAVs still need to be examined in mixed traffic scenarios. To fully realize the benefits of CAVs at intersections with mixed traffic flow, this study proposes an innovative approach known as the mixed platoon and intersection coordination strategy. Specifically, the mixed platoon and intersection coordination strategy refers to constructing a formation area and a speed planning area at the intersection, achieving mixed multi-vehicle formation and right-of-way conflict resolution. However, the mixed traffic consisting of CAV and HDV is a partially controllable system. Therefore, we propose a bi-level control framework where the upper-level determines the right-of-way for traffic demands from different directions while the lower-level implements speed planning for CAVs. To efficiently assess the available crossing span for vehicles and calculate their acceleration, we propose a mixed traffic resource allocation algorithm in conjunction with a reinforcement learning-based speed planning algorithm. Simulation experiments demonstrate that the proposed strategy can rapidly identify available crossing spans for vehicles and guide them to the intersection in their target states. Compared to traditional intersection control methods, the proposed approach enhances traffic system efficiency, increasing effectiveness as the penetration rate of CAVs rises.