基于强化学习的工业机器人操作器自归巢能力探究  

Exploration of Self-Homing Ability of Industrial Robot Manipulators Based on Reinforcement Learning

作　　者：刘云[1] Liu Yun(Jiangxi Vocational College of Mechanical and Electrical Technology,Nanchang 330013,Jiangxi Province,China)

机构地区：[1]江西机电职业技术学院,江西南昌330013

出　　处：《科学与信息化》2022年第24期112-114,共3页Technology and Information

摘　　要：文章提出了一种基于非视觉、强化学习的并行代理设置方法,使工业机器人细胞具有自应能力,消除了机器人操纵器手动编程的需要。这种方法假设可以感知到状态定义中预先编码的未知机器人细胞环境,以便学习到的策略可以在没有进一步训练的情况下转移。该方法已在模拟工业机器人单元中得到了验证,并证明了在模拟中得到的策略可转移到相应的真实工业机器人单元中,并可推广到制造环境中的其他机器人系统。This paper proposes a parallel proxy setting method based on non-visualization and reinforcement learning,which enables self-responsive ability of industrial robot cells and eliminates the need for manual programming of robot manipulators.This approach assumes that the unknown robot cell environment pre-coded in the state definition can be perceived,so that the learned strategy can be transferred without further training.This method has been validated in simulated industrial robot elements,and it has demonstrated that the strategies obtained in the simulation can be transferred to the corresponding real industrial robot elements and generalized to other robot systems in the manufacturing environment.

关 键 词：强化学习 自归巢 并行代理 工业机器人机械手 

分 类 号：TP2[自动化与计算机技术—检测技术与自动化装置]