Autonomous Vehicle Optimal Path Following Control with Robustness and Passenger Comfort

In this paper, a model predictive path following control system is proposed for an autonomous vehicle. First a dynamic-kinematic model for the non-holonomic robot is derived. While simple and straightforward, it can effectively represent the transient motion and instantaneous errors, with the desired reference path taken into account. Based on the model, model predictive controller (MPC) is designed to yield optimal behavior, with constraints taken into consideration. In addition to physical constraints, limitations as to passenger comfort are also regarded. The optimal feature of MPC is exploited to find optimal trajectories from initial errors to the desired path. The control system is well capable of path tracking and double lane change. As will be seen, thanks to online receding horizon, the control system yields margins of robustness for parametric uncertainties.