Trajectory Tracking Control of Unmanned Vehicle Based on Data-driven Optimization

Since the traditional Model Predictive Control(MPC), which is widely used for trajectory tracking of autonomous vehicle, cannot analyze and determine specific parameters of the controller by mathematical methods. In this paper, a trajectory tracking control method based on model prediction is proposed to solve the problem of unmanned vehicle trajectory tracking, and the controller is optimized in a performance objective driven way. Specifically, the cost function of the model predictive controller is parameterized. And the global optimal performance in a specific scenario as the goal to build the global performance cost function. Then, the global performance cost is expressed as a Gaussian process, and new parameters of the next optimization are inferred by Bayesian optimization. The controller parameters of global performance optimization are found with a small learning cost through multiple iterations to improve tracking performance. To verify the effectiveness of this data-driven optimization algorithm, lane-changing experiments with Carsim and Matlab/Simulink are carried out. According to the test data, it is proved that the performance of trajectory tracking under this data-driven MPC algorithm is optimized.