Cycle-Removal-Based Priority Policies Coordination for Distributed Intelligent Intersection Management

Priority policies, which schedule vehicles' passing order in non-signalized intersection management systems, have been a critical task since the development of connected and autonomous vehicles. However, current priority policies lack the ability to coordinate with each other in the same intersection, leaving little flexibility to handle the complex real-world scenarios such as unexpected emergencies, special requirements from manufacturers or service providers, non-uniform computing environments, and faulty control. In this work, a cycle-removalbased coordination strategy is proposed to resolve the deadlock created by the incompatibility between different priority policies. This strategy involves solving an optimization problem—a constrained version of the minimum feedback arc set problem. This newly proposed problem builds upon established algorithmic foundations for the unconstrained version but requires significant adaptations to meet the unique constraints and demands of practical distributed intersection management. Particularly, an exact method and a heuristic are studied and adapted with numerical experiments designed to evaluate their performance. The experimental results show their effectiveness in achieving our goals, which enables the coordination of priority policies and improves an intersection management system's capability for handling complicated requirements.