LS-SVM application for ship course model predictive control

Since ship dynamics have the characteristics of high non-linearization, great inertia and time-variability, the LS-SVM with RBF kernel function, due to its outstanding performance to approximate non-linear function models, is used in this paper to make effective detection of ship model. The obtained non-linear approximation model is then combined with model predictive control to derive system dynamic model based on LS-SVM. The linearized analytical control method is furthermore derived to achieve ship course prediction and control, based on real time linearization system model. It is demonstrated by computer simulations that LS-SVM exhibits faster calculation speed by reducing algorithm complexity and shows excellent generalization capability for small samples. The proposed ship course predictive control system demonstrates acceptable immunity against external disturbance and parameter perturbation with satisfactory controllability.