现代制造工程 ›› 2024, Vol. 521 ›› Issue (2): 31-37.doi: 10.16731/j.cnki.1671-3133.2024.02.005

• 机器人技术 • 上一篇    下一篇

基于机器学习的工业机器人多目标轨迹规划*

张学聪1, 晁永生1, 李纯艳1, 周江林2   

  1. 1 新疆大学机械工程学院,乌鲁木齐 830017;
    2 新疆大学交通运输工程学院,乌鲁木齐 830017
  • 收稿日期:2023-03-17 出版日期:2024-02-18 发布日期:2024-05-29
  • 作者简介:张学聪,硕士研究生,主要研究方向为机器人轨迹规划。E-mail:zxc22314@163.com;晁永生,博士,副教授,硕士生导师,主要研究方向为智能制造。E-mail:cys21st@163.com
  • 基金资助:
    *新疆维吾尔自治区自然基金项目(2022D01C37)

Multi-objective trajectory planning for industrial robots based on machine learning

ZHANG Xuecong1, CHAO Yongsheng1, LI Chunyan1, ZHOU Jianglin2   

  1. 1 School of Mechanical Engineering,Xinjiang University,Urumqi 830017,China;
    2 School of Transportation Engineering,Xinjiang University,Urumqi 830017,China
  • Received:2023-03-17 Online:2024-02-18 Published:2024-05-29

摘要: 为满足机器人工作轨迹的多样化需求,以时间、能量及冲击为优化目标提出了一种新的多目标麻雀搜索算法,用于寻找机器人的最优轨迹。首先,通过7次B样条插值方法构造关节空间轨迹,以此确立多目标综合最优轨迹规划模型。其次,采用违反约束度计算、非支配排序以及精英保留来改进麻雀搜索算法,使其能够处理机器人多目标轨迹规划问题。最后,用袋装树分类算法对随机种群内数据进行了筛选,并搭建5层BP神经网络来替代改进多目标麻雀搜索算法中适应度值的数值计算部分,从而提高算法求解效率。通过MATLAB仿真与实验证明了该算法优化所得轨迹的可行性及有效性。

关键词: 工业机器人, 轨迹规划, 机器学习, 麻雀搜索算法, 多目标优化

Abstract: In order to meet the diverse needs of robot work trajectories, a new multi-objective sparrow search algorithm with optimization objectives of time,energy and impact was proposed to find the optimal trajectory of robots.Firstly,the joint space trajectory was constructed by the 7-degree B-spline interpolation method,and the multi-objective synthetic optimal trajectory planning model was established.Secondly,the sparrow search algorithm was improved by constraint violation computing,non-dominated ranking and elitist retention to solve the multi-objective trajectory planning problem.Finally,the data in the random population was deleted by the bag-tree classification algorithm,and a five-layer BP neural network was built to replace and improve the numerical calculation part of the fitness value in the multi-objective sparrow search algorithm,thus improving the efficiency of the algorithm.Through the MATLAB simulation and experiment,the feasibility and validity of the algorithm were demonstrated.

Key words: industrial robot, trajectory planning, machine learning, sparrow search algorithm, multi-objective optimization

中图分类号: 

版权所有 © 《现代制造工程》编辑部 
地址:北京市东城区东四块玉南街28号 邮编:100061 电话:010-67126028 电子信箱:2645173083@qq.com
本系统由北京玛格泰克科技发展有限公司设计开发 技术支持:support@magtech.com.cn