CISCAV: Consensus-based Intersection Scheduling for Connected Autonomous Vehicles

Intelligent vehicles that autonomously plan, communicate, and perform intersection crossings will reduce accidents and delays compared to human drivers. Previously proposed solutions require the use of additional expensive infrastructure, such as centralized Intersection Managers, or are based on theoretical control optimization with over-optimistic predictability of each vehicle’s behavior.In this paper, we propose a decentralized intersection management that requires no additional infrastructure and can tolerate timing deviations due to unpredictable but detectable events, such as pedestrian movement or ambulances. Vehicles cooperate via direct VANET communication to agree on a schedule that specifies the groups and the order in which approaching vehicles will cross the intersection.We have implemented our protocol as simulation using Artery and SUMO. Our experiments show that CISCAV ensures safety and reduces the average delay in high traffic conditions to 32 s compared to 150 s for conventional crossing policies such as traffic lights or priority roads.