Unscented Kalman Filter trained neural networks based rudder roll stabilization system for ship in waves

Abstract

The large roll motion of ships sailing in the seaway is undesirable because it may lead to the seasickness of crew and unsafety of vessels and cargoes, thus it needs to be reduced. The aim of this study is to design a rudder roll stabilization system based on Radial Basis Function Neural Network (RBFNN) control algorithm for ship advancing in the seaway only through rudder actions. In the proposed stabilization system, the course keeping controller and the roll damping controller were accomplished by utilizing modified Unscented Kalman Filter (UKF) training algorithm, and implemented in parallel to maintain the orientation and reduce roll motion simultaneously. The nonlinear mathematical model, which includes manoeuvring characteristics and wave disturbances, was adopted to analyse ship’s responses. Various sailing states and the external wave disturbances were considered to validate the performance and robustness of the proposed roll stabilizer. The results indicate that the designed control system performs better than the Back Propagation (BP) neural networks based control system and conventional Proportional-Derivative (PD) based control system in terms of reducing roll motion for ship in waves.