Comparison of Path Tracking and Torque-Vectoring Controllers for Autonomous Electric Vehicles

Abstract

Steering control for path tracking in autonomous vehicles is well documented in the literature. Also, continuous direct yaw moment control, i.e., torque-vectoring, applied to human-driven electric vehicles with multiple motors is extensively researched. However, the combination of both controllers is not yet well understood. This paper analyzes the benefits of torque-vectoring in an autonomous electric vehicle, either by integrating the torque-vectoring system into the path tracking controller, or through its separate implementation alongside the steering controller for path tracking. A selection of path tracking controllers is compared in obstacle avoidance tests simulated with an experimentally validated vehicle dynamics model. A genetic optimization is used to select the controller parameters. Simulation results confirm that torque-vectoring is beneficial to autonomous vehicle response. The integrated controllers achieve the best performance if they are tuned for the specific tire-road friction condition. However, they can also cause unstable behavior when they operate under lower friction conditions without any re-tuning. On the other hand, separate torque-vectoring implementations provide a consistently stable cornering response for a wide range of friction conditions. Controllers with preview formulations, or based on appropriate reference paths with respect to the middle line of the available lane, are beneficial to the path tracking performance.