不确定检测环境下强化学习覆盖路径规划研究
A Coverage Path Planning Method with Reinforcement Learning Considering Manufacturing Process Uncertainty
李彦征 1刘银华 1赵文政 1孙芮2
作者信息
- 1. 上海理工大学 机械工程学院,上海 200093
- 2. 上海交通大学 机械与动力工程学院,上海 200240
- 折叠
摘要
机器人扫描测量系统在汽车质量检测领域获得广泛应用,尤其数模环境下基于仿生优化算法的视点采样与路径规划研究取得较大进展.然而,基于名义数模环境的路径规划结果难以适用实际不确定性检测环境.为此,本文提出基于改进的蒙特卡洛树搜索的视点自适应采样方法,在线生成工业机器人运动轨迹.通过车门内板案例的对比分析,验证本文方法的有效性,为实现不确定制造环境下工艺路径的在线规划提供理论依据.
Abstract
The robotic scanning system has been widely used in the quality inspection field of automobiles,especially the studies of viewpoint sampling and path planning based on the genetic optimization algorithm in the model-based environment.However,the path planning results based on the nominal models are difficult to apply to the actual inspection environment.To address this problem,a viewpoint adaptive sampling method is proposed based on an improved Monte Carlo tree search,and industrial robot motion trajectories are planned online.Finally,the case of the car door inner panel was used to illustrate the effectiveness of the method.
关键词
光学检测/覆盖路径规划/制造误差/运动规划Key words
optical inspection/coverage path planning/manufacturing deviation/motion planning引用本文复制引用
基金项目
国家自然科学基金项目(51875362)
上海市自然科学基金项目(21ZR1444500)
机械系统与振动国家重点实验开放基金项目(MSV202010)
出版年
2024
国家科技期刊平台
NSTL
维普
万方数据