A Reinforcement Learning Algorithm for Underwater Environment Search

Known as an intelligent learning algorithm that can be employed with- out any sufficient prior information, the reinforcement learning algorithm has been widely used recently. The underwater unknown 3D environment search is a typical application scenario lacking any prior knowledge. This study proposes an algorithm framework including one agent and an intelligent swarm based on the idea of reinforcement learning. This learning algorithm can be adopted to train only one individual. Many of such individuals will then create an intelligent group which can efficiently and completely search underwater unknown 3D environments. Moreover, the proposed algorithm needs no further prior knowledge of the unknown environment. Even if the algorithm fails to reach the optimal level in the use of reinforcement learning, the results confirm the applicability prospect of reinforcement learning in this field and provide an appealing direction.