Injury robot gait self-learning fusing biological inspiration and deep reinforcement learning

Abstract

The invention provides an injury state robot gait self-learning method fusing biological inspiration and deep reinforcement learning. The method comprises the following steps: modeling a hexapod robot and constructing a reinforcement learning control system; setting a state value and an action value; setting a reward function; constructing an Actor-Critic neural network; selecting a reinforcement learning algorithm for network parameter optimization; carrying out bionic feature constraint on the hexapod robot model; and training the gait of the hexapod robot in typical injury states. According to the method, the gait of the robot is generated by using the method of fusing biological inspiration and deep reinforcement learning, so that the robot can be subjected to gait adjustment through gait self-learning after being damaged, and the method has important significance in improving the survivability of the robot in a complex environment.

Description

Technical field[0001]The present invention relates to a marginal approach to a non-injury state of six-foot robotic, especially a bio-inspiration and depth reinforcement of a damage state robotic analysis.Background technique[0002]The bionic six-foot robot has a rich skill, strong environmental adaptability, especially suitable for search and rescue, reconnaissance and material transport in harsh field environments. The robots operating in complex hazardous environments are very susceptible to various damage, resulting in different degrees of damage. Normally, the movements and control capabilities of robots will be rapidly weakened. If the damaged robot is adjusted and learned in a short period of time, this is essential for damage to the damaged robot self-rescue and enhanced living capacity.[0003]When using a model-based traditional method for robot control, developers first need to perform kinematics and kinetics modeling, and programs for a specific task in accordance with the ...

AI Technical Summary

Problems solved by technology

This method undoubtedly has shortcomings. Once the robot is damaged and its own state changes, the original robot model will no longer be applicable, and developers cannot pre-create a robot model that covers all damage types of the robot.

Images