Multi-agent Signalless Intersection Management with Dynamic Platoon Formation

We propose a novel mechanism to manage platoons of autonomous vehicles at traffic intersections. Our mechanism optimises the formation forming vehicle platoons to minimises overall waiting time, allowing the optimal platoon size to be determined dynamically, thus minimising overall travel time. In addition, we introduce a conflict resolution algorithm, which dynamically authorises multiple platoons to manoeuvre even when the majority are single vehicles. Our empirical evaluation shows that, for single intersections, our mechanism can reduce the average travel time by up to ~65% compared to conventional fixed-time traffic signals and up to ~4% compared to advanced non-platoon-based signal-less first-come-first-served approaches. Moreover, from the corridor-level aspect, our mechanism can reduce the weighted average trip duration up to ~22% compared to the fixed-time traffic signals and up to ~45% compared to the signal-less first-come-first-served approaches.