Hierarchical Motion Planning for Consistent and Safe Decisions in Cooperative Autonomous Driving

The immersion of autonomous cars in continuously changing environments of on-road traffic requires procedures for decision-making with fast adaptation as well as guarantees on safe motion and collision-avoidance. This contribution proposes a three-layer hierarchic decomposition of the task of automatically steering the autonomous car along a designated route in cooperation with neighbored vehicles. The upper layer of the hierarchy identifies cooperative groups of those vehicles which are involved in a joint scenario for a phase of the planning horizon. The medium layer employs set-based computations of the free space for any vehicle of a joint scenario together with constrained optimal control to determine optimized motion plans. These plans are used on the lower layer as reference signals for tracking control in order to realize motion trajectories. The architecture ensures consistency of the vehicle motion with respect to safety for given assumptions, as well as relatively small computation times by combining offline with online computation.