Path Tracking Controllers for Fast Skidding Rover

Abstract

This article presents a comparative analysis between two new path tracking controllers, a linear and non-linear one. First, dynamic model of the vehicle moving on a horizontal plane is developed and validated with the experimental platform Spido. Then, to reduce the complexity, the model is linearized by assuming that the side slip angles remain small and a linear lateral slippage model introducing the lateral contact stiffness is used. The kinematics of the vehicle is also linearized by assuming that the rover remains close a reference path and with a small relative heading angle. The linearized model permits the design of a Linear Quadratic Regulator LQR that minimizing the angle deviation and the lateral error. Then, nonlinear predictive controller is presented. This approach combines the kinematic model of the path-tracking with the dynamic model of the vehicle to deduce a non-linear Continuous Generalized Predictive Controller NCGPC controller. The predictive approach is dedicated to achieve an accurate path tracking and minimizing the angle deviation and the lateral error during prediction horizon time. Finally, we compare performance results between the NCGPC and the LQR controller. The simulation results show that both controllers have equivalent path tracking performance only in the straight line sections. However, predictive controller clearly outperform LQR controller when cornering. NCGPC controller has a higher rate of convergence to the reference trajectory with minimal lateral position error, thanks to the anticipation of future reference path changes.