Adaptive critic-based neuro-fuzzy controller for dynamic position of ships

In this paper, three adaptive critic-based neuro-fuzzy controllers are presented for improving attitude and position control of ships. The controllers include vessel position and heading errors and their derivatives as inputs. Three critic based reinforcement learning methods evaluate the situations of the ship in terms of satisfactory achievements of the control goals. The critic agent output, namely the reinforcement signal, is a measure of the controlled system stress. The controller modifies its characteristics in a way that the critic stress is decreased. The proposed controller has a simple structure and shows satisfactory transient responses and robustness to model uncertainty. (C) 2015 Sharif University of Technology. All rights reserved.