Contingency Planning for Automated Vehicles in Urban Traffic

Abstract

Imagine a world where everyone could get from point to point in an easy and safely way. Where elderly or even visually impaired would greatly benefit from autonomous driving. The Automotive Department of Institut für Verkehrssystemtechnik (Institute of Transportation Systems) of the Deutsches Zentrum für Luft- und Raumfahrt e.V. (DLR) Braunschweig is conducting research in the field of cooperative and automated vehicles driving in urban traffic. It is intended to develop a contingency planner for urban traffic that finds a solution in a 0.1[s] time frame. Anytime Search algorithms can be utilized in constrained contingency planning to compute a safe and feasible path in a free environment by taking advantage of an inflated heuristic and pre-computed motion primitives. In the offline phase of the contingency planner it is intended to develop a set of motion primitives, following a nonlinear system model, bounded by physical constraints which takes the burden of feasibility verification from the online phase, where time has a higher cost. The online phase is the search phase in which a graph of maneuver segments is constructed using the motion primitives. An improvement in the number of explored invalid nodes and runtime was verified in the experiments. In the first a reduction of twenty to sixty times and in the second it was accomplished a six to ten times faster search. The strategy implemented, where the inflated part of the overestimated heuristic could be utilized to make a better informed search, gives a framework for future adaptations.