Deep reinforcement learning dual-arm robot control method based on demonstration examples, and system

Abstract

The invention relates to a deep reinforcement learning dual-arm robot control method based on demonstration examples, and a system. The deep reinforcement learning dual-arm robot control method based on the demonstration examples comprises the steps of obtaining the three-dimensional coordinate motion increment of the position of a clamping jaw at the tail end of each group of mechanical arm based on the shortest distance between an initial coordinate of the clamping jaw at the tail end of a dual-arm robot and an initial coordinate of a target object, and forming a first demonstration example used for demonstration; after the clamping jaw closes to grab the object, obtaining the three-dimensional coordinate motion increment of the position of the clamping jaw at the tail end of each group of mechanical arm based on the shortest distance between the initial coordinate of the target object and a final coordinate of the target object, and forming a second demonstration example used for demonstration; and utilizing the demonstration examples of the two parts to form a complete expert demonstration track, adopting an experience playback pool of a double-agent deep deterministic policy gradient algorithm to randomly extract a small batch of samples to form training data, training a neural network, and achieving control over the dual-arm robot. The reinforcement learning algorithm reinforces the action sequence of the demonstration examples, and fast learning is carried out until the task is completed.

Description

technical field [0001] The invention relates to the field of robot control, in particular to a demonstration example-based deep reinforcement learning dual-arm robot control method and system. Background technique [0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art. [0003] The grasping task of the robot is a multi-step sequential decision-making task in the continuous action space, which needs to execute multiple steps continuously within a period of time, and any problem in any step will lead to the failure of the task. A dual-arm robot is a robot that completes the grasping task through two sets of robotic arms. There is a certain motion constraint relationship between the grasping arms, and it is not a simple superposition of two single-arm robots. [0004] The coordinated control of the dual-arm robot means that the two robotic arms adapt to the environment and coope...

AI Technical Summary

Problems solved by technology

In the prior art, only relying on the reinforcement learning algorithm to control the dual-arm robot, the robot arms as the intelligent body cannot receive positive rewards and cannot carry out effective learning, thus making the coordination control effect of the dual arms unsatisfactory and easy There is a problem that only one robot arm completes the grabbing action and the other robot arm does not move

Images