On-line modular identification based on recursive PLS regression with application to predictive ship motion control

An on-line modular predictor is proposed for control application of ship maneuvering motion. This approach combines parametric identification with non-parametric identifications, which are realized based upon recursive partial least squares regression and variable neural network, respectively. Simulation of predictive ship course control is performed by employing the modular predictor for on-line motion prediction. Simulations results of ship motion prediction and control demonstrate the feasibility and effectiveness of the proposed modular predictor.