Reinforcement learning-based motion planning of a triangular floating platform under environmental disturbances

摘要

This paper investigates the use of reinforcement learning for the motion planing of an autonomous triangular marine platform in unknown environments under various environmental disturbances. The marine platform is over-actuated, i.e. it has more control inputs than degrees of freedom. The proposed approach uses an online least-squared policy iteration scheme for value function approximation in order to estimate optimal policy. We evaluate our approach in simulation, taking under consideration the dynamics of the platform, the dynamics and limitations of the actuators, under the presence of wind, and sea current disturbances. We report simulation results concerning its performance on estimating optimal navigation policies to unknown environments. Despite the model dynamics, the actuation dynamics and limitations, and the environmental disturbances, the presented results are promising.