Deep reinforcement learning-based complex terrain self-adaptive motion control method for hexapod robot

Abstract

The invention provides a deep reinforcement learning-based complex terrain self-adaptive motion control method for a hexapod robot. Compared with a conventional mode of realizing motion control of therobot via beforehand programming, the method disclosed in the invention is advantageous in that deep reinforcement learning is used to allow the robot to adaptively adjust motion strategies accordingto complex changes of environment, and "survival rates" and adaptability in complex environments can be improved; compared with a single actor-critic deep reinforcement learning framework, the methodis advantageous in that a multiple actor-multiple critic deep reinforcement learning framework is realized, and training can be sped up and a neural network can be converged fast during deep neural network training operation.

Description

technical field [0001] The present invention relates to the field of machine learning and the technical field of robot motion control, and more specifically, to a complex terrain adaptive motion control method for a hexapod robot based on deep reinforcement learning. Background technique [0002] With the development of the times, mobile robots play an increasingly important role in real life and play a role in various fields, such as search and rescue, disaster response, medical care, transportation, etc. The real world is an unstructured and dynamic environment, and robot motion control is a multi-dimensional control problem. Creating a robot that can adaptively move in such an environment has always been a major challenge in the field of robotics. [0003] Traditional hexapod robot motion control methods are pre-programmed, that is, encoding various motion gaits of the hexapod robot into the motion control chip of the hexapod robot. However, in the environment with compl...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to solve the technical problem of the hexapod robot that is prone to fall to the ground due to unstable center of gravity when the hexapod robot is in a complex environment or its own leg structure has problems in the prior art, and provides a method based on Adaptive motion control method for complex terrain of hexapod robot based on deep reinforcement learning

Images