Robot motion parameter adaptive control method and system based on deep reinforcement learning

• Main Authors: YANG YA, ZHANG ZHIPENG, MA BAOPING, LI XIAOQIANG, WANG CHUNLEI, SHAO HAICUN, REN LIANG, PENG CHANGWU
• Format: Patent
• Language: Chinese; English
• Published: 08.10.2021
• Subjects: CHAMBERS PROVIDED WITH MANIPULATION DEVICES; HAND TOOLS; MANIPULATORS; PERFORMING OPERATIONS; PORTABLE POWER-DRIVEN TOOLS; TRANSPORTING

The invention provides a robot motion parameter adaptive control method and system based on deep reinforcement learning. The method comprises the steps that an intelligent agent is constructed in a simulation environment, wherein the intelligent agent comprises a strategy neural network, a value neural network and a task planning module; based on guided reinforcement learning, the strategy neural network in the intelligent agent is trained according to sample parameters; based on layered reinforcement learning, strategy promotion and strategy evaluation are alternately performed on the strategy neural network and the value neural network in the intelligent agent in sequence according to multiple sub-tasks and corresponding reward functions, and a trained strategy neural network model is obtained; and based on the trained strategy neural network model, a control parameter optimization value is output to a controller according to a target task, and therefore the controller controls a robot according to the cont