Multi-Hypothesis Evasive Maneuvers for Safe Autonomous Navigation

Abstract

Recent advances in Autonomous Vehicles (AV) driving raised up all the importance to ensure the complete reliability of AV maneuvers even in highly dynamic and uncertain environments/situations. To reach this purpose, autonomous vehicles operating in such complex environments need methods which generalize to unpredictable situations. Validating the safety of self-driving vehicles can prove the coherence of the vehicles' behavior, reduce remaining risks and the need for extensive testing and, more importantly, allow us to plan evasive maneuver. This paper proposes a multi-hypothesis evasive strategy able to cope with any dynamic traffic situation. It is based on: a Sequential Decision Networks for Maneuver Selection and Verification (SDN-MSV) that calculates discrete evasive action maneuver based on defined situational criteria; an exhaustive evasive trajectory generation that considers multi-hypothesis kinematic and dynamic configuration; and on a multi-criteria optimization algorithm able to generate the corresponding low-level control that allows the ego-vehicle to pursue the advised collision-free evasive maneuver. At the same time, jerk is minimized while punishing high acceleration and curvature rate to provide enhanced comfort for passengers. Several simulation results show the good performance of the overall proposed evasive strategy.