Application of Reinforcement Learning to Generate Non-linear Optimal Feedback Controller for Ship's Automatic Berthing System
Naoki Mizuno, Tatsuya Koide
Abstract
In this paper, we present an application of reinforcement learning for automatic ship's berthing system. In the proposed method, the model-based reinforcement learning is used to generate non-linear feedback controller. The proposed system is composed of an Actor-Critic algorithm suitable for continuous state and a function approximator by Radial Basis Function networks. To evaluate the performance of the proposed system, extensive computer simulations and actual sea tests are carried out using small training ship Shioji-Maru III under various conditions. As a result, we can see that the proposed non-linear optimal feedback controller by reinforcement learning is useful for designing automatic berthing system for the ship.