An optimal ship autopilot using a multivariate auto-regressive exogenous model

Linear Quadratic Gaussian (LQG) control algorithm linked to Recursive Least Squares (RLS) algorithm has been applied to a multivariate auto-regressive exogenous (MARX) model of ships to construct a ship autopilot. It has been found from computer simulation and full-scale experiments aboard a training ship that the autopilot is robust and has good features in both course keeping and course changing.