ROBOT MOTION PARAMETER ADAPTIVE CONTROL METHOD AND SYSTEM BASED ON DEEP REINFORCEMENT LEARNING

WIPO Patent Application WO/2022/223056

A1

A robot motion parameter adaptive control method and system based on deep reinforcement learning. The robot motion parameter adaptive control method based on deep reinforcement learning comprises: constructing an agent in a simulation environment, wherein the agent comprises a policy neural network, a value neural network and a task planning module; on the basis of guided reinforcement learning, training the policy neural network in the agent according to a sample parameter; on the basis of hierarchical reinforcement learning, sequentially and alternately performing policy improvement and policy evaluation on the policy neural network and the value neural network in the agent according to a plurality of sub-tasks and reward functions corresponding thereto, so as to obtain a completely trained policy neural network model; and on the basis of the completely trained policy neural network model, outputting a control parameter optimization value to a controller according to a target task, such that the controller controls a robot according to the control parameter optimization value.

REN LIANG (CN)
WANG CHUNLEI (CN)
YANG YA (CN)
SHAO HAICUN (CN)
ZHANG ZHIPENG (CN)
MA BAOPING (CN)
PENG CHANGWU (CN)
LI XIAOQIANG (CN)

PCT/CN2022/104735

October 27, 2022

July 08, 2022

Click for automatic bibliography generation  

THE 21TH RESEARCH INSTITUTE OF CHINA ELECTRONIC TECH GROUP CORPORATION (CN)

B25J9/16; B62D57/032; G05D1/02

CN113478486A	2021-10-08
CN110861084A	2020-03-06
CN111208822A	2020-05-29
CN111580385A	2020-08-25
US10786900B1	2020-09-29
US20180157973A1	2018-06-07

BEIJING WUZHOUYANGHE & PARTNERS (CN)

View/Download PDF      PDF Help

Previous Patent: TARGET ASSEMBLY AND X-RAY MICROSCOPE

Next Patent: INTELLIGENT OPERATION AND MAINTENANCE TEST PLATFORM FOR ACCELERATED FAILURE EXCITATION OF INDUSTRIAL...