Intelligent Motion Control to Enhance the Swimming Performance of a Biomimetic Underwater Vehicle Using Reinforcement Learning Approach

This article develops an intelligent motion control strategy using reinforcement learning to regulate a biomimetic autonomous underwater vehicle (BAUV) swimming performance. The BAUV is driven by the oscillatory motion of a lunate-shaped caudal fin. To navigate effectively, the vehicle must regulate its propeller’s motion to swim toward goals. The caudal fin’s oscillatory motion is defined by sine functions, where the amplitude, frequency, bias, and time shifting are the main flapping parameters that must be regulated. The developed algorithm, based on the n-step state-action-reward-state-action on-policy learning, fine-tunes these parameters. Further, deep Q learning network speed adjusters were designed to enhance BAUV guidance. By integrating waypoint guidance systems and intelligent path trackers, the vehicle reduces its heading deviation and distance to goals. The effectiveness of these methods was validated through simulations featuring various disturbances like longitudinal, lateral, and swirl currents. The proposed scheme allowed the vehicle to reach desired targets efficiently, even in the presence of strong currents.