Nonlinear Identification of Ship Autopilot Models

Abstract

The purpose of this thesis has been to develop methods for identification of nonlinear models to be used in ship autopilots. An accurate model is essential when developing autopilot systems. Although a number of identification methods are available, only a few ship maneuvers are described in the literature. During this report a literary study on nonlinear identification methods has been carried out and an overview over several methods is presented. A new maneuvering model derived by Andrew Ross is simulated to generate measurement data. Based on the measurements during several predefined maneuvers, an iterative prediction error method is applied to identify the parameters of two different autopilot models. Secondly, a new ship maneuver is suggested for identification of ship steering dynamics. Compared to the classic turning circle and zig-zag maneuver the new maneuver shows better convergence properties and perform good adaptation of the dynamics. At last the identified autopilot models are verified by simulating the ship in closed-loop using a model-based autopilot controller.