Gaussian Process Based Robust Trajectory Tracking of Autonomous Underwater Vehicle
Yongxu He, Yuxin Zhao, Geng Xu, Xiong Deng · 2021 China Automation Congress (CAC) · 2021
This paper develops a novel robust control method for trajectory tracking of autonomous underwater vehicles by addressing significant modeling uncertainties and external disturbances. Given a prior dynamic model, we apply sparse online Gaussian process technique to acquire nonparametric estimates of the unknown dynamics along with their probabilistic confidence intervals. The employed Gaussian process regression approach has a predefined limited budget for the training data set, enabling computation efficiency and online learning. By explicitly incorporating Gaussian process posterior means and variances into a sliding mode control framework, the proposed method guarantees asymptotic stabilization of a vehicle to the desired trajectory with a specific probability. Simulation results demonstrate the effectiveness of the proposed control method.