ANALYSIS ON THE LOCOMOTION OF CABLE TUNNEL INSPECTION QUADRUPED ROBOT BASED ON DEEP REINFORCEMENT LEARNING

The quadruped robot applying the expert skill learning system can learn and generate adaptive skills from a group of representative expert skills, screen and master these combined skills through deep reinforcement learning, so as to select different skills in different environments to move in stranger environment. This approach leverages the advantages of trained expert skills and the fast online synthesis of adaptive policies to generate responsive motor skills during the changing tasks. The cable tunnel unmanned inspection quadruped robot is equipped with a 5-DOF mechanical arm with a sensor module on the top to realize the cable condition detection. The combination of the intelligent sensors can make the robot obtain the ability of environment perception similar to human beings, so that the quadruped robot can complete the instructions safely and accurately in the cable tunnel environment, collect and analyse the environment and cable condition parameters, find problems and give feedback in time.