Automated Vehicles in Hazardous Merging Traffic: A Chance-Constrained Approach

Abstract

Abstract To realize the energy efficiency and productivity benefits of automated driving, control algorithms must function safely among conventional vehicles. Prototype tests on public roads have revealed a trend of human-driven vehicles rear-ending automated ones. A popular belief holds that unusually conservative control algorithms play a role in such collisions. In August 2018, an automated vehicle was rear-ended while waiting to merge. Inspired by that incident, this paper examines a resemblant scenario in simulation using model predictive control for the automated vehicle. Constraint setup alternatives to avoid collisions with inattentive following vehicles are proposed and assessed in this simulation environment. In one variant, imminent rear-end collisions are detected and constraints are modified to promote more aggressive merging during such an event. Results show that higher-performing chance constraint designs can reduce collision probability, but may have other adverse effects depending on the particular algorithm.