A new ship's auto pilot design through a stochastic model

A new type of ship's autopilot system is designed by a statistical approach. A ship's motion at sea is described by a multi-variable autoregressive model using minimum AIC (Akaike's Information Criterion) procedure. Through the fitted model, the ship's behavior is analyzed and an optimal control law for a ship under a newly introduced criterion function is derived. The feasibility of our control system is checked by both digital and hybrid simulations. The results of the simulation show that with our controller the yaw motion is depressed through smoother and less rudder motion than that of the conventional autopilot systems and the ill effect of rolling is avoided. It is expected that the controller has another merit: it is quite robust for possible changes of external environments. Finally, a successful result of an actual sea test is briefly discussed. Thus, the possibility of realizing an entirely new autopilot system by a stochastic model is demonstrated.