This paper treats of the applications of the pole assignment self-tuning control theory to the identification and design of ship's auto-pilot systems. An on-line estimation algorithm known as the Recursive Least Squares Algorithm is used to identify parameters of the control system. Two typical types of the pole assignment self-tuning control algorithms that can be linked to the on-line estimation algorithm to form self-tuning control systems for ships are introduced. The authors discuss how to choose the control parameters by the rules of thumb in the actual sea trials. The pole assignment self-tuning controllers designed to steer ships are verified by experiments aboard the training ship Shioji Maru of Tokyo University of Mercantile Marine. The aboard experiments result in the confirmation that it is possible to keep and to change the ship's course as desired by the designed controllers.