Design of a Reinforcement Learning-Based Speed Compensator for Unmanned Aerial Vehicle in Complex Environments

Due to the complexity of the marine environment and the uncertainty of ship movements, altitude control of UAV is particularly important when approaching and landing on the deck of a ship. This paper focuses on unmanned helicopters as its research subject. Conventional altitude control systems may have difficulty in ensuring fast and stable landings under certain extreme conditions. Therefore, designing a new UAV altitude control method that can adapt to complex sea conditions has become a current problem to be solved. Designing a reinforcement learning based rotational speed compensator for UAV as a redundant controller to optimise UAV altitude control performance for the above problem. The compensator is capable of adjusting the UAV’s rotational speed in real time to compensate for altitude deviations due to external environmental disturbances and the UAV’s own dynamic characteristics. By introducing reinforcement learning algorithms, especially the DDPG algorithm, this compensator is able to learn the optimal RPM adjustment strategy in a continuous trial-and-error process, which improves the UAV’s rapidity and stability during the landing process.