AUV Position Tracking Control Using End-to-End Deep Reinforcement Learning

CARLUCHO, IGNACIO; DE PAULA, MARIANO; WANG, S.; PETILLOT, YVAN; ACOSTA, GERARDO G.

Congreso; MTS/IEEE OCEANS 2018 - Charleston; 2018

In this article we consider the navigation problemfor an autonomous underwater vehicle (AUV) for reachinga desired way-point. The navigation problem in underwatervehicles presents major problems, the highly coupled dynamicsof the vehicles and the unknown parameters of the dynamicmodel, make the need for complex control architectures. However,current developments in reinforcement learning show promisingresults for robotics applications. In particular underwaterautonomous vehicles could benefit from this new techniques,achieving adaptive behavior for real-time problem solving. Basedon this developments the navigation problem is solved usingdeep reinforcement learning, in particular the deep deterministicpolicy gradient. In this proposal a model free approach is used,where the raw sensor information is used as inputs to a policynetwork, and the outputs of this network are directly mapped tothe thrusters. In addition an adaptive goal driven architecture isused to allow the agent to reach variable way points consistently.The obtained simulated results show its capacity for successfullysolving AUV navigation problems.