Comparison of Linear and Nonlinear Model Predictive Control for Vehicle Path Following

This technical paper investigates and compares the performance of two model predictive control algorithms employed in the control of a ground vehicle. Both predictive controllers are based on a simplified bicycle model of the vehicle. The main difference is that one controller uses a nonlinear version of the bicycle model and the other one further simplifies it by linearizing it around the operating point. The study centers around evaluating the controllers’ effectiveness in a dynamic scenario through the execution of a double-lane change maneuver. By assessing trajectory tracking and computational efficiency, this paper aims to delineate the impact of linearization choices on the controllers’ overall performance. Both controllers are extensively tested in realistic simulations on a full non-linear multi-body model of the vehicle sourced from the CommonRoad Python library. The results of these simulations provide insights into the tradeoffs between model accuracy and computational complexity.