Target Following for an Autonomous Underwater Vehicle Using Regularized ELM-based Reinforcement Learning

In order to achieve autonomy and intelligence of autonomous underwater vehicle (AUV), dynamic target following in the unknown environment is one of the important problems to solve. Reinforcement learning (RL) offers the possibility of learning a policy to solve a particular task without manual intervention and previous experience. However, RL algorithms are not competent for continuous space problems existing universally in underwater environment. In this paper, the method of an AUV using regularized extreme learning machine (RELM) based RL is proposed for following the moving target. The most typical and common method of RL used in applications is Q learning, which can generate the proper policy for AUV control. And RELM is proposed as a modification of ELM, which can guarantee the generalization performance and work with continuous states and actions. The simulation results have shown that AUV successfully tracks and follows the moving target by the proposed method.