Design Of Control System Of Tracked Mobile Robot Based On Reinforcement Learning

With the development of science and technology,robots have gradually entered various fields of human production and life.Among them,mobile robots have greatly expanded the scope of robot operations and increased flexibility.Compared with other types of mobile robots,tracked mobile robots have the advantages of adapting to all-terrain,have strong towing ability work in dangerous and harsh environments,and have many applications in agriculture,military and other fields.They have always been a focus of attention at home and abroad..This paper takes the tracked mobile robot as the research object,based on the reinforcement learning theory,to study its control problems in unknown and uncertain environments.The main tasks completed are as follows:(1)We construct related models of the tracked mobile robot control system,including the kinematics model and dynamics model of the tracked mobile robot.It also conducts theoretical derivation and analysis of the posture transformation,movement and force conditions during the construction of the tracked mobile robot control system.(2)To improve Q learning,we propose an incremental discretization method,and further use neural network to map the discrete state action space to the continuous state action space,give a Q learning controller based on the continuous action space.Considering that the tracked mobile robot is disturbed by the outside world during the movement,a robust reinforcement learning optimization controller is designed to ensure the stability and optimization of the tracked mobile robot during the movement.The simulation experiments were carried out respectively,and the simulation experiment results were analyzed to verify the effectiveness of the controller.(3)We build a tracked-type mobile robot experimental platform,and carry out robot motion control experiments on the experimental platform.First,we build a ROS-based tracked mobile robot software system framework to realize the functions of each node.Then,we build the hardware system of the tracked mobile robot,and introduce the selection of hardware and its functions in detail.Finally,the algorithm proposed in this paper is used on the experimental platform to conduct a tracked mobile robot control experiment.By designing a variety of motion trajectories to verify the effectiveness of the algorithm through comparative analysis.