Path following control using robust sliding mode for an autonomous surface vessel based on ε- support vector regression identified steering model

Abstract

Due to the complexity and unmodeled dynamics along with internal model uncertainties, an ε- Support Vector Regression (SVR) algorithm is implemented to determine the parameters of Nomoto's steering equation of an Autonomous Surface Vessel (ASV) based on the experimental data for 20°−20° zigzag manoeuver. An analytical formulation is used to compare and check the accuracy of the system identification method. Then, considering the uncertainty in estimated parameters, a sliding mode controller for the path following control is implemented with −20%,0%, and 20% uncertainty in the obtained coefficients, (K, T). To follow the desired path, the Dynamic Line of Sight algorithm is proposed and its parameters are obtained based on the optimal performance of the ASV in converging to the desired path. The results demonstrate that theε-SVR can adequately estimate the hydrodynamic coefficients using free-running model tests. Similarly, the designed controller also showed that ASV followed the predefined path with an average following error of 9.5%. The simulation results considering the bounded of 20% uncertainty in hydrodynamic parameters also show that the uncertainty effects are less on the path following control (average of −9.9% in −20% and 2.8% in +20% uncertainty) and not made instability in the designed controller.