Bayesian Optimization based Trajectory Tracking Controller for Steer-by-Wire Vehicles

Abstract

In the intelligent driving system of wire-controlled vehicles, the vehicle trajectory tracking controller is important in improving driving experience and reducing traffic accident rate. To make the vehicle trajectory tracking controller’s adaptability improved when facing different stability states, the study uses model predictive control to construct a vehicle trajectory tracking controller and combines the Gaussian Mixture Model-Hidden Markov model to construct a vehicle stability grading criterion and then uses Bayesian optimization to improve the controller parameters to design the improved vehicle trajectory tracking controller. The test results show that the trajectory tracking controller designed in this research deviates from the reference trajectory by 0.1m under the speed conditions of 60km/h and 80km/h. The overall fluctuation of transverse velocity is -0.1m/s to 0.1m/s and -0.3m/s to 0.3m/s and the angular velocity of the transverse pendulum is 0.3rad/s. The peak of positive and negative transverse error is 0.16m and 0.25m with 0.2m and 0.25m. The RMSE is 0.0892 and 0.0990. The vehicle drops to low-risk level 1 when its stability reaches high-risk levels 3 and 4. The results show that the Bayesian optimization-based wire steering vehicle trajectory tracking controller can improve the adaptability and trajectory tracking accuracy of different stability states and ensure the vehicle driving stability.