Manoeuvring of underwater snake robot with tail thrust using the actor-critic neural network super-twisting sliding mode control in the uncertain environment and disturbances

Snake robots are used for surveillance in unknown underwater environments. The robot comes across the drag force and added mass effect during the motion. The speed of the robot is very low in the underwater environment. Therefore, some thrust force is required to increase the speed. In this work, the snake robot follows a path in uncertain environments with external disturbances. The speed of the snake robot is improved by implementing a thruster at the tail of a snake robot. The equation of motion for the snake robot is derived by considering the effect of the thrust force on each link. The virtual holonomic constraints are formulated to derive the reduced-order dynamical system. The control objective is defined based on reduced-order dynamics using the sliding surface approach. The super-twisting sliding mode control (STSMC) scheme with reinforcement learning (i.e. actor-critic neural network) is proposed for the motion control of the snake robot in the uncertain underwater environment. The dynamic model of the snake robot (equivalent control law) is estimated using the reinforcement learning control approach. The switching control law is designed using the STSMC. The proposed control scheme not only improves the tracking errors but also reduces the chattering effect and control effort. Finally, these results are verified by comparing them to the existing control scheme. Overall, the online dynamic model estimation problem is addressed by designing the controller for the head link of the snake robot.