Stochastic Optimal Approach to the Steering of an Autonomous Vehicle through a Sequence of Roadways

This paper discusses the implementation of a stochastic optimal controller for steering a vehicle to robustly follow an unpredictably winding road. The controller is based on a bicycle model and the road is defined as a sequence of roadways. Each roadway has a fixed position, but its orientation is uncertain. To anticipate this uncertainty, we model the orientation with a Brownian stochastic process, which serves as a stochastic process model for the orientation observations. The stochastic controller based on such a model implicitly creates a robust road following controller. The control design is illustrated with numerical simulations and implemented for steering a car in a high-fidelity car simulator.