首页|基于积分强化学习的助力型人-机协同搬运研究

基于积分强化学习的助力型人-机协同搬运研究

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
日常实践中存在大量人类与人类之间的繁重物件转移和搬运操作,人们需要共同将此类物体笨拙地搬运至目标位置,这都需要耗费大量的时间和人力.随着协作机器人的广泛应用,这些简单、耗时又耗力的任务,都可让协作机器人作为人类的同事,以自然又高效的方式无缝地协同工作,就像人类与人类一样.为此,针对人-机协同搬运中位置精度低,所需操作者交互力过高等问题,提出了一种基于积分强化学习的助力型人-机协同搬运方法.根据人-机协同搬运的动力学模型,重点分析协同搬运的全流程.提出协同搬运的内外环数学模型,然后采用基于积分强化学习的自适应阻抗控制,来实现人与机器人的协同搬运.最后通过人-机协同搬运的实验,验证了所提出方法的有效性.
Force-Supported Human-Robot Collaborative Handling Technology Using Integral Reinforcement Learning
IIn daily practice,humans frequently transfer and handle heavy objects.Moving heavy items to target positions can be item-consuming and require significant manpower.With the increasing use of collaborative robots(Cobots),the Cobots can serve as human colleagues in a natural and efficient manner to work together seamlessly to complete the simple,time-consuming and labor-intensive tasks,just as humans working together.To address the low positioning accuracy and high interaction force,this paper proposes an integral reinforcement learning-based method for force-supported human-robot collaborative handling.The dynamic model of human-robot collaboration is used to analyze the entire handling process.Mathematical models for the inner and outer loops of collaborative handling are put forward,and integral reinforcement learning and adaptive impedance control are integrated to establish a method for collaborative handling between humans and robots.Experimental verification is conducted to show that this method meets daily handling requirements.The results of the human-robot collaborative handling experiment show that the proposed method is effective.

human-robot collaborationobject handlingimpedance controlintegral reinforcement learning

齐春雨、张树忠、陈旭飞、王馨裕、张弓

展开 >

福建理工大学 福建省智能加工技术及装备重点实验室,福州 350108

广州先进技术研究所前瞻科学与技术研究中心,广州 511458

人-机协同 物体搬运 阻抗控制 积分强化学习

国家自然科学基金广东省自然科学基金福建省智能加工技术及装备重点实验室开放基金福建省科技创新重点项目福建省中央引导地方科技发展资金项目(2022)

620730922021A1515012638KF-01-220052022G020072022L3014

2024

机械设计与研究
上海交通大学

机械设计与研究

CSTPCD北大核心
影响因子：0.531
ISSN：1006-2343
年,卷(期)：2024.40(1)
  • 19