Model predictive attitude control of unmanned surface vehicle based on short-time wave prediction

Unmanned Surface Vehicles (USVs) are widely used in commercial and marine scientific research. In high sea conditions, USV is susceptible to significant random wave disturbances, necessitating the adoption of attitude control mechanisms to ensure safe operation. Existing control algorithms, mostly relying on Proportional Integral and Differential (PID) control and other common methods, tend to concentrate solely on the current model state, and seldom predict the future evolution of the model in combination with the state of the waves. This limitation hinders the optimization of attitude control in rough seas. To address this challenge, this paper proposes a model predictive attitude control approach grounded in Long Short-Term Memory neural network (LSTM) wave forecasting. By forecasting the wave disturbances on the USV in the coming period of time horizon and optimizing maneuvering commands, it identifies the optimal thruster actuation combination. This strategy dynamically adapts to sea conditions in real-time, enhancing the precision and accuracy of attitude control. The stability of this control strategy is validated using the discrete Lyapunov stability criterion. Simulation results demonstrate that our proposed control strategy exhibits remarkable roll and pitch reduction performance, surpassing PID control, linear quadratic regulator (LQR) control, model predictive control (MPC) without disturbance compensation, and MPC with disturbance compensation. Moreover, it maintains a superior control effect even in the presence of state feedback noise, indicating promising application prospects.