摘要
作为自动化和智能化时代的代表,机器人技术的发展成为智能控制领域研究的焦点,各种基于机器人的智能控制技术应运而生,机器人被越来越多地应用于实现与环境之间的复杂多接触交互任务.本文以机器人复杂多接触交互任务为核心问题展开讨论,结合基于强化学习的机器人智能体训练相关研究,对基于强化学习方法实现机器人多接触交互任务展开综述.概述了强化学习在机器人多接触任务研究中的代表性研究,当前研究中存在的问题以及改进多接触交互任务实验效果的优化方法,结合当前研究成果和各优化方法特点对未来机器人多接触交互任务的智能控制方法进行了展望.
As a representative of the automation and intelligence era,robot technology has become the focus of the intelligent control research,Various robot intelligent control technologies have been developed,Robots are more and more used to achieve some complex contact-rich interaction tasks.This paper chooses robot complex contact-rich interaction tasks as a topic,combining the research of robot reinforcement learning,to make a survey of research about the robot contact-rich interaction tasks based on reinforcement learning.We review some representative researches of implementing reinforcement learning in robot contact-rich interaction tasks,analyze the existing problems in these researches,summarize the optimization methods of solving these problems to improve the experimental effects,and finally make a prospect on the future of robot contact-rich tasks.
作者
柯恺宸
金士博
高博扬
黄行蓉
Ke Kaichen;Jin Shibo;Gao Boyang;Huang Xingrong(Sino-French Engineering School,Beihang University,Beijing 100190,China;Hangzhou Innovation Institute(Yuhang),Beihang University,Hangzhou 310023,China;Faculty of Computing,Harbin Institute of Technology,Harbin 150001,China;Research Institute of Aero Engine,Beihang University,Beijing 100190,China)
出处
《动力学与控制学报》
2023年第12期53-69,共17页
Journal of Dynamics and Control
基金
国家自然科学基金资助项目(52105083,52175071,U2341231)
两机基础科学中心国际合作项目(P2022-C-III-001-001)。
关键词
强化学习
智能控制
机器人
多接触交互任务
reinforcement learning
intelligent control
robot
contact-rich tasks
