Robust multistage nonlinear model predictive control on an autonomous marine surface vehicle

Due to the need for a higher safety level, autonomous marine surface vehicles have been increasingly applied for various applications in recent years. Nonlinear model predictive control (NMPC) has increasingly become a more preferred control technique. However, it has still limited use in industry due to its high computational load which is mainly caused by its complex Newton numerical method and Hessian matrix evaluations. In this study, we have developed a robust multistage NMPC algorithm to accurately control an autonomous marine surface vehicle when model uncertainties or environmental disturbances are present. The robust multistage NMPC is compared with the standard NMPC and perfect information NMPC. Advanced symbolic computation approach has been used to implement all three NMPC programs, where the complex gradient vector, Hessian matrix, and nonlinear dynamic optimization problem are derived analytically and evaluated with machine accuracy. The simulation results reveal that the robust multistage NMPC approach is successfully able to better control an autonomous marine surface vehicle than the standard NMPC method, and impressively is almost as good as the perfect information NMPC when model uncertainties or environmental disturbances are present. Furthermore, it has been demonstrated that the robust NMPC is also very effective for a large scaled-up prototype system.