Learning Control for Autonomous Driving on Slippery Snowy Road Conditions

This paper presents an investigation of Gaussian Processes (GPs) in combination with model predictive control (MPC) for autonomous driving control on slippery snowy road conditions. A double lane change scenario with two different road friction coefficients is considered for learning the GP model. The model is then incorporated into the MPC algorithm development. The performance of the GP-MPC controller is evaluated and compared with conventional MPC controller. The validation is conducted based on a co-simulation platform that simulates high fidelity vehicle/tire dynamics and snowy traffic environment in different setting conditions, respectively. The results demonstrate that the GP-MPC controller can achieve better trajectory tracking performance and with less control input than the conventional MPC controller however with higher computation time.