A novel sequential learning algorithm for RBF networks and its application to ship predictive control

A radial basis function (RBF) network-based predictive control strategy is proposed for ship control. The RBF network is on-line trained to identify the time-varying system dynamics using a novel sequential learning algorithm referred to as dynamic orthogonal structure adaptation (DOSA) algorithm. The combination of neural network identification and predictive control mechanism minimizes the effects of ship's time-varying dynamics and long time delay, enables accurate and smooth control of ship under various disturbances and random noises. Simulation results of ship track-keeping control demonstrate the applicability and effectiveness of the control strategy. The quick and adaptive learning algorithm gives RBF network more representing abilities to model nonlinear systems with unstable or unknown dynamics.