[1]陈富强,陈振庭,许丽娟.基于改进深度强化学习的工业机器人多拐点避障[J].机械与电子,2025,(08):61-66.
　CHEN Fuqiang,CHEN Zhenting,XU Lijuan.Multi-inflection Point Obstacle Avoidance for Industrial Robots Based on Improved Deep Reinforcement Learning[J].Machinery & Electronics,2025,(08):61-66.
点击复制

基于改进深度强化学习的工业机器人多拐点避障()
分享到：

《机械与电子》[ISSN:1001-2257/CN:52-1052/TH]

卷:
期数:
2025年08期
页码:
61-66
栏目:
智能制造
出版日期:
2025-08-25

文章信息/Info

Title:
Multi-inflection Point Obstacle Avoidance for Industrial Robots Based on Improved Deep Reinforcement Learning
文章编号:
1001-2257 ( 2025 ) 08-0061-06
作者:
陈富强; 陈振庭; 许丽娟
广州华商学院人工智能学院,广东 广州 511300
Author(s):
CHEN Fuqiang ;  CHEN Zhenting ;  XU Lijuan
( School of Artificial Intelligence , Guangzhou Huashang College , Guangzhou 511300 , China )
关键词:
改进深度强化学习; 工业机器人; 避障控制; 离散空间; 引导奖赏函数
Keywords:
mproved deep reinforcement learning ;  industrial robot ;  obstacle avoidance control ;  discrete space ;  guided reward function
分类号:
TP242.2
文献标志码:
A
摘要:
针对工业机器人在具有先验信息的拐角障碍环境中自主导航时,未考虑与障碍物距离最优性,导致在障碍物多拐点处存在冗余路径及深度学习过程反复试错的问题,提出一种改进深度强化学习的工业机器人避障控制方法。通过分析机器人与障碍物在坐标空间中的横、纵坐标差值,考虑静动态障碍物距离差值占比建立引导奖赏函数,根据距离变化动态调整奖惩值优化避障策略,以避障距离奖惩值为最优距离建立离散空间改进算法并给出最优控制函数。实验结果表明,在多拐点环境中所提方法避障控制效果佳,能在最短时间内实现精准避障,控制性能优异,具有实用价值。
Abstract:
A deep reinforcement learning based obstacle avoidance control method for industrial robots is proposed to address the problem of redundant paths and repeated trial and error in the deep learning process due to the lack of optimal distance from obstacles when autonomously navigating in a corner obstacle environment with prior information.By analyzing the difference between the horizontal and vertical coordinates of the robot and obstacles in the coordinate space , considering the proportion of the distance difference between static and dynamic obstacles , a guided reward function is established.The reward and punishment values are dynamically adjusted according to the distance changes to optimize the obstacle avoidance strategy.An improved algorithm is developed by establishing a discrete space based on obstacle avoidance distance reward and punishment values as the optimal distance , and the corresponding optimal control function is derived.The experimental results show that this method has good obstacle avoidance control effect in multi-inflection point environments and can achieve accurate obstacle avoidance in the shortest time.It has excellent control performance , and has practical value.

参考文献/References:

[ 1 ] LI R G , WU H N.Multi-robot source location of scalar fields by a novel swarm search mechanism with collision / obstacle avoidance [ J ] .IEEE Transactions on intelligent transportation systems , 2022 , 23 ( 1 ): 249-264.

[ 2 ] 庄红超,王柠,董凯伦,等 . 非完整约束大负重比六足机器人多机动态协同编队避障控制策略[ J ] . 机械工程学报,2024 , 60 ( 1 ): 284-295.
[ 3 ] LU K , DAI S L , JIN X.Adaptive angle-constrained enclosing control for multirobot systems using bearing measurements [ J ] .IEEE Transactions on automatic control , 2024 , 69 ( 2 ): 1324-1331.
[ 4 ] 赵海文,罗元铭,张雅丽,等 . 约束空间工业机器人姿态搜索及避障研究[ J ] . 组合机床与自动化加工技术,2024 ( 5 ): 77-81.
[ 5 ] SUN Y , YUAN Q , GAO Q , et al.A multiple environment available path planning based on an improved A* algorithm [ J ] .International journal of computational intelligence systems , 2024 , 17 ( 1 ): 172.
[ 6 ] 张涛,陈璋,李玉梅,等 . 融合改进 A* 算法与动态窗口法的机器人避障研究[ J ] . 仪表技术与传感器,2023( 4 ): 102-106.
[ 7 ] 王康康,桂宏凡 . 基于改进麻雀搜索算法的外骨骼机器人步态检测[ J ] . 电子器件, 2023 , 46 ( 2 ): 423-429.
[ 8 ] 张颖,乔贵方,王保升,等 . 基于优化位姿集的工业机器人运动学参数辨识方法研究[ J ] . 中国测试,2023 , 49( 3 ): 91-95 , 103.
[ 9 ] 周伟,潘金宝,王林琳,等 . 基于改进鲸鱼算法和 A* 算法的地面放线机器人路径规划[ J ] . 现代制造工程,2023 ( 12 ): 68-75 , 83.
[ 10 ] 胡钊政,伍锦祥,肖汉彪,等 . 基于逆投影差分的移动机器人障碍物快速检测[ J ] . 哈尔滨工业大学学报,2022 , 54 ( 11 ): 95-102.
[ 11 ] 吴涛,谢志军,陈科伟 . 改进 D* lite 和时间弹性带法的移动机器人路径规划[ J ] . 传感技术学报, 2022 , 35( 4 ): 486-494.
[ 12 ] 欧阳云,高振国,范丽玲,等 . 采用 RSPM-PS 算法的机械手末端避障路径规划[ J ] . 华侨大学学报(自然科学版),2023 , 44 ( 3 ): 290-300.
[ 13 ] 孙立香,孙晓娴,刘成菊,等 . 人群环境中基于深度强化学习的移动机器人避障算法 [ J ] . 信息与控制,2022 , 51 ( 1 ): 107-118.
[ 14 ] 安燕霞,郑晓霞 . 基于分层强化学习的机器人自主避障算法仿真[ J ] . 计算机仿真, 2024 , 41 ( 4 ): 397-401.
[ 15 ] 杨芳,席瑾 . 基于双目视觉和 SIFT 算法的羽毛球接球机器人避障研究[ J ] . 自动化与仪器仪表, 2024 ( 2 ):213-217 , 223.
[ 16 ] 罗丽霞,常金勇 . 二维未知环境下移动机器人二次规划避障路径控制[ J ] . 机械设计与研究, 2024 , 40 ( 1 ):97-101 , 108.
[ 17 ] 张涛,陈璋,李玉梅,等 . 融合改进 A*算法与动态窗口法的机器人避障研究[ J ] . 仪表技术与传感器, 2023( 4 ): 102-106.

相似文献/References:

[1]孔民秀,赵 宁.机器人示教臂系统的示教实现[J].机械与电子,2015,(10):76.
　KONG Minxiu,ZHAO Ning.Realization of Teaching Method by Robot Teaching Arm[J].Machinery & Electronics,2015,(08):76.
[2]陈世钟,刘延遂,吴品弘,等.基于刚度性能的机器人臂长优化[J].机械与电子,2015,(06):67.
　CHEN Shizhong,LIU Yansui,WU Pinhong,et al.Arm Link Length Optimization of Robots Based on Stiffness Performance[J].Machinery & Electronics,2015,(08):67.
[3]刘 爽.工业机器人在物流拣选场景的应用[J].机械与电子,2019,(11):71.
　Application of Industria Robots in Logistics Picking Scenarios[J].Machinery & Electronics,2019,(08):71.
[4]张 华.基于非线性优化算法的工业机器人轨迹跟踪自动控制[J].机械与电子,2023,41(04):55.
　ZHANG Hua.Automatic Trajectory Tracking Control of Industrial Robot Based on Nonlinear Optimization Algorithm[J].Machinery & Electronics,2023,41(08):55.
[5]梁存仙,焦建静,赵志鹏,等.基于改进遗传算法的工业机器人视觉动态分拣方法研究[J].机械与电子,2025,(03):60.
　LIANG Cunxian,JIAO Jianjing,ZHAO Zhipeng,et al.Research on Visual Dynamic Sorting Method of Industrial Robot Based on Improved Genetic Algorithm[J].Machinery & Electronics,2025,(08):60.

备注/Memo

备注/Memo:
收稿日期: 2024-12-05
基金项目:广州华商学院校内导师制科研基金资助项目( 2024HSDS12 )
作者简介:陈富强 ( 1995- ),男,湖北黄冈人,硕士,讲师,研究方向为云计算、图像处理和信息安全;陈振庭 ( 1993- ),男,广东汕头人,硕士,助教,研究方向为计算机视觉、数据分析与挖掘;许丽娟 ( 1979- ),女,广东广州人,硕士,副教授,研究方向为图像处理、数据分析,通信作者, E-mail : d104544@163.com 。
更新日期/Last Update: 2025-09-05