Research on Control of Underwater Robotic Arm Based on Unknown Gap Lag and Reinforcement Learning

Yunlong Bai
School of Automation, Harbin University of Science &Technology, Harbin150080, Heilongjiang, China
Yu Cao
School of Automation, Harbin University of Science &Technology, Harbin150080, Heilongjiang, China
Hongsen Zhang
School of Automation, Harbin University of Science &Technology, Harbin150080, Heilongjiang, China
Jiaxing Tong
School of Automation, Harbin University of Science &Technology, Harbin150080, Heilongjiang, China
Bo Hu
School of Automation, Harbin University of Science &Technology, Harbin150080, Heilongjiang, China
Yu Zhao
School of Automation, Harbin University of Science &Technology, Harbin150080, Heilongjiang, China
Mingyu Li
School of Automation, Harbin University of Science &Technology, Harbin150080, Heilongjiang, China

Abstract:

With the depletion of land resources, more and more people turn their attention to the urgent exploration of Marine resources. The application of various underwater platforms equipped with underwater robotic arms for underwater operations is one of the mainstream development trends of future Marine development tasks. Due to the complex and changeable underwater operating environment, it is difficult to establish accurate dynamic models for underwater robotic arms, and the application effect of traditional control algorithms can no longer meet the increasing needs of underwater operations. As a machine learning method that does not depend on the problem model, deep reinforcement learning algorithm can optimize the strategy by learning experience through interaction with the environment. Due to the influence of water pressure, temperature and other factors, the gap lag will occur when the underwater robot arm performs the task. This hysteresis will reduce the control accuracy of the robot arm and even cause the robot arm to lose control. Therefore, the research on the control method of underwater robot arm based on unknown gap lag and reinforcement learning has important theoretical significance and practical application value. In this paper, a series of control strategies are studied based on reinforcement learning algorithms to solve the unknown gap lag problem of underwater robotic arms. Firstly, an inverse adaptive controller design scheme for uncertain nonlinear single-input single-output systems with unknown gap hysteresis is proposed. Secondly, based on reinforcement learning algorithm, a control algorithm suitable for underwater robot arms are designed. Finally, the effectiveness and practicability of the proposed control strategy are verified by simulation experiments. The research results of this paper provide theoretical basis and technical support for the intelligent control of underwater robot arm and contribute to the further development of underwater robot technology.