Optimization-Based Maneuver Automata for Cooperative Trajectory Planning of Autonomous Vehicles

In this paper, a major challenge in the field of trajectory planning for autonomous on-road vehicles is addressed: to calculate trajectories efficiently and reliably. The proposed method is based on a cooperative maneuver automaton, in which each maneuver is formulated as a discrete-time hybrid automaton that defines constraints on the behavior of a group of autonomous vehicles. For trajectory planning, these constraints enter an optimal control problem, which is formulated as a Mixed-Integer Quadratic Program (MIQP). Despite good practical performance, these programs have adverse worst-case run-time properties, which, however, are improved by the proposed approach. Also, calculation of controllable sets for the hybrid automata allows to determine both the feasibility of an optimal control problem for a given initial state and the required time to complete the maneuver in advance- without solving the optimization problem. The efficacy of the proposed method is demonstrated in an example scenario.