Contingency Planning for Automated Vehicles

Abstract

Automated vehicles face several challenges when planning their actions in an uncertain and partially known urban environment. Creating optimal and feasible paths in those scenarios in a reduced runtime is a difficult problem. A planner adapted to a cluster of scenarios and integrated in an hierarchical planning system with other modules is a viable solution. Such a structure would require a behavioural planner to share computational resources between the planners. A contingency planner was developed for emergency situations in urban traffic, with the aim to compute a safe maneuver in a 0.1[s] time frame. In this paper, we present a continuing anytime based search algorithm, which allows the heuristic to be sensitive to the environment while preserving sub-optimally bound solution and completeness properties. It is provided simulation results for an emergency situation to demonstrate the efficiency of the proposed approach. An environment sensitive heuristic allows a better informed search, which we believe could be adaptable to a wide range of scenarios.