Robust hybrid control for autonomous vehicle motion planning

Abstract

The operation of an autonomous vehicle in an unknown, dynamic environment is a very complex problem, especially when the vehicle is required to use its full maneuvering capabilities, and to react in real time to changes in the operational environment. A possible approach to reduce the computational complexity of the motion planning problem for a nonlinear, high dimensional system, is based on a quantization of the system dynamics, leading to a control architecture based on a hybrid automaton, the states of which represent feasible trajectory primitives for the vehicle. The paper focuses on the feasibility of this approach, in the presence of disturbances and uncertainties in the plant and/or in the environment: the structure of a robust hybrid automaton is defined and its properties are analyzed. In particular, we address the issues of well-posedness, consistency and reachability. For the case of autonomous vehicles, we provide sufficient conditions to guarantee reachability of the automaton.