Ship's optimal autopilot with a multivariate auto-regressive exogenous model

This paper presents a new application of the linear quadratic gaussian (LQG) control algorithm linked to the recursive least squares (RLS) algorithm applied to a multivariate auto-regressive exogenous (MARX) model of ships to construct an autopilot for steering ships. Simulation performed for a training ship is described. As the first step of designing; a tracking system, the optimal autopilot with the MARX model was used to both keep and change the ship's course during full-scale experiments aboard the training ship. It has been found that the autopilot is robust and has good performance for steering ships. Copyright (C) 2000 IFAC.