Reachability based Online Safety Verification for High-Density Urban Air Mobility Trajectory Planning

Abstract

This paper presents a safe and scalable real-time trajectory planning framework for high-density Urban Air Mobility (UAM). The framework extends our previously developed highly efficient Markov Decision Process (MDP) based trajectory planner by adopting a correct-by-construction approach to verify the safety of planning actions. The trajectory planner works in a decentralized manner that allows each aircraft to generate its safe trajectory by accounting for the reachable sets of nearby aircraft. The proposed safety verification module employs a highly scalable data-driven reachability analysis tool to ensure collision-free trajectory planning. Furthermore, the utilized tool over-approximates the reachable set of each aircraft using a discrepancy function, which it learns online from simulation traces. We finally demonstrate the efficacy of the proposed trajectory planner with simulation experiments of up to 120 aircraft in a UAM setting.

Authors

• Abenezer Taye
• Josh Bertram

Keywords

• Engineering
• Computer Science
• Environmental Science