Evasion planning for autonomous intersection control based on an optimized conflict point control formulation

Abstract Autonomous intersection management (AIM) has been widely researched, but previous studies assume that vehicles will follow assigned trajectories precisely. The purpose of this paper is to investigate the safety buffers needed between intersecting vehicles to avoid a collision if a vehicle malfunctions. We optimize vehicle trajectories by deciding the arrival times at each conflict point (point of possible intersection with other vehicles) along each vehicle’s trajectory. Because intersecting vehicles rely on the intersection manager (IM) to detect and communicate malfunctions, the reaction time from the IM determines the minimum safety buffer needed. Although a smaller reaction time reduces the safety buffer, it increases the probability that the IM falsely detects a malfunction, instructing vehicles to stop and creating unnecessary delays. This paper develops a mathematical safety buffer for intersecting vehicles, linearizes this time separation, and constructs a combined mixed-integer linear program. A complete protocol is presented and simulated for normal circumstances, emergency circumstances, and recovery circumstances. Sensitivity analyses on various reaction times show the tradeoff between low reaction times (more false positives) and high reaction times (greater safety buffer).