Intersection Deadlock Protocol on Mixed Autonomous Vehicles

During rush hours, intersection deadlock frequently occurs when vehicles inside the intersection halt and wait for spaces occupied by competing streams to be released. This waiting behavior forms a loop structure, requiring human intervention for coordination. With the increasing prevalence of connected and autonomous vehicles (CAVs), once CAVs become trapped in a deadlock, they may remain indefinitely stuck. Existing research on CAVs primarily focuses on intersection efficiency, largely disregarding the deadlock phenomenon. This chapter proposes an intersection deadlock protocol for mixed autonomous flow, comprising a cooperative deadlock detection and a deadlock avoidance/unlock protocol. Two types of deadlock, weak deadlock and strong deadlock, are considered based on fixed steering angles. Weak deadlock detection employs an escaping distance propagation algorithm, while strong deadlock detection utilizes a dynamic programming-like method. Deadlock avoidance is achieved through a chain-spillover-free and loop-free strategy, and the unlock protocol employs rerouting and backward-forward strategies. The proposed protocol is validated using the Carla platform, demonstrating a capacity improvement of up to 35% and enabling deadlock detection 6 s earlier before its occurrence. The average unlocking time for a single deadlock is approximately 6 s.