Gliding Control of Underwater Gliding Snake-like Robot based on Reinforcement Learning

The control of the gliding action of underwater gliding snake-like robot is mainly studied in this paper. In order to solve the problem that the hydrodynamic environment is hard to model, the method of reinforcement learning is used to make the underwater gliding snake-like robot adapt to the underwater environment and automatically learn the gliding actions. A modified Monte Carlo policy gradient algorithm using pre-processed neural network is proposed to solve the problem that the states of the robot are difficult to be observed completely due to its complex structure. The gliding control problem of the underwater gliding snake-like robot can be approximated as a Markov Decision Process, so as to obtain an effective gliding control policy. Simulation results show the effectiveness of the proposed method.