Efficient Mixed Traffic Intersection Management under Uncertainty

Abstract

Despite the extensive amount of research that exists regarding intersection crossing of Connected and Autonomous Vehicles, implementation of most of the already proposed theoretical algorithms is currently not realistic due to either their computational complexity, or strong assumptions about traffic data acquisition. An intermediary step would be to consider mixed traffic. Motivated by these facts, we study an unsignalized intersection crossing problem with mixed traffic and develop an online, safety-critical optimization scheme under localization and detection uncertainties. The centralized controller follows an Enhanced First-In-First-Out reservation policy and with the use of sequential optimization it ensures no more than one vehicle is present inside the protected zone of the intersection in order to avoid collisions, even when noisy measurements are present. Comparisons are made between different control and protected zone sizes and conclusions are drawn about the impact of uncertainties on the efficiency and robustness of the system.