An On-line Learning Algorithm for RBF Networks and its Application to Ship Inverse Control

An on-line learning algorithm is introduced referred to as dynamic orthogonal structure adaptation (DOSA) algorithm for constructing radial basis function (RBF) networks with variable network structure. The RBF network is on-line adapted for both network structure and connecting parameters. Based on DOSA algorithm, an inverse control strategy is proposed and applied to ship control. Simulation results of ship course control experiment demonstrate the applicability and effectiveness of the proposed inverse control strategy.