Deep Learning for Station Keeping of AUVs

Abstract

Control of underwater vehicles remains an active research topic within the literature. Multiple challenges exists for controlling an underwater vehicle, including highly nonlinear effects due to hydrodynamics. Control based models seek to model the underlying dynamics but suffer from the balance between tractable computation and performance. Machine Learning (ML) control techniques show promise as an alternative to classical model-based approaches. This article investigates the application of a model-free deep reinforcement learning algorithm, Deep Deterministic Policy Gradient (DDPG), for station keeping in six degrees of freedom (DOF) for an underwater vehicle.