Reinforcement learning for autonomous underwater vehicles (AUVs): navigating challenges in dynamic and energy-constrained environments

Abstract

Autonomous Underwater Vehicles (AUVs) are essential for underwater exploration, inspection, and environmental surveillance. Nevertheless, navigation, obstacle avoidance, and energy efficiency are greatly hindered by the ever-changing underwater environments. Reinforcement Learning (RL) has arisen as a revolutionary method for tackling these challenges. This paper examines significant progress in reinforcement learning algorithms, emphasizing their application in the training of autonomous underwater vehicles in both simulated and real-world environments. The review synthesizes findings from multiple studies, identifies gaps in existing research, and highlights the potential of algorithms such as Deep Deterministic Policy Gradient (DDPG) for continuous control tasks. This review offers an extensive examination of current methodologies, their constraints, and avenues for future investigation. Received: 02 November 2024 Accepted: 22 December 2024 Published: 24 December 2024