Path Planning for Autonomous Underwater Vehicles in Uncertain Environments

Abstract

This paper addresses the path planning problem for autonomous underwater vehicles (AUVs) in uncertain ocean environments. In comparison with mobile robots, the motion of an AUV is frequently interfered by ocean currents. Thus, the path planning solution should not only consider generating low-cost, collision-free and dynamically feasible paths but also dealing with environmental disturbances and actuation errors. In this work, the novel feature of the proposed solution is that a reinforcement learning approach is presented, and the exploration strategies for optimizing the leaning episodes are also evaluated and discussed. The proposed solution is validated by a simulated tested constructed upon a grid-based map with a model of an AUV in uncertain environments.