机械臂环境三维重建与避障算法研究  

Research on Environmental 3D Reconstruction and Obstacle Avoidance Algorithm of Industrial Robots

作　　者：陈新度[1,3] 徐学 高萌 孔德良 CHEN Xin-du;XU Xue;GAO Meng;KONG De-iang(School of Electromechanical Engineering Guangdong University of Technology,Guangdong Guangzhou 510006,China;Foshan Institute of Intelligent Equipment Technology,Guangdong Foshan 528234,China;State Key Labora-tory of Precision Electronic Manufacturing Technology and Equipment,Guangdong Guangzhou 510006,China)

机构地区：[1]广东工业大学机电工程学院,广东广州510006 [2]佛山智能装备技术研究院,广东佛山528234 [3]广东工业大学省部共建精密电子制造技术与装备国家重点实验室,广东广州510006

出　　处：《机械设计与制造》2024年第1期347-352,358,共7页Machinery Design & Manufacture

基　　金：2019年佛山市核心技术攻关项目—新一代智能工业机器人研发与应用(19200010011367)。

摘　　要：为了解决机器人在未知环境中的避障规划难题,提出了一种基于视觉的三维重建与避障规划方法。通过深度相机(RealSense D435i)与机械臂组成手眼系统,首先,根据目标物体与ArUco码的相对关系,通过相机获取的ArUco标记上的特征点坐标,求解出目标物姿态;然后,改进三维重建方法,通过融合机械臂工具末端位姿和多帧点云数据,对环境进行较高精度的三维重建,作为后续规划的障碍空间;最后,应用了快速扩展随机树(RRT)的改进算法进行机械臂的避障运动规划。研究搭建了仿真及控制平台,来进行验证。实验表明:环境三维建模的精度维持在8mm以内,避障成功率为96.5%,平均规划时间为1.2s,验证了方法的可行性。Aiming at solving the problem of obstacle avoidance and motion planning for robots in unknown environments,a method of 3D reconstruction and obstacle avoidance for industrial robot arm based on visual was researched.The hand-eye sys-tem consisted of the depth camera(RealSense D435i)and the manipulator.Firstly,the posture of the target object was calculated by using the coordinate of the feature point on the ArUco mark obtained by the depth camera.Then,the 3D reconstruction method is improved.By fusing the end pose of the manipulator and multi-frame point cloud data,the environment is reconstructed with higher precision in three dimensions as the obstacle space for subsequent planning.Finally,the improved algorithm of Rapidly Expanding Random Tree(RRT)was applied to the obstacle avoidance and motion planning of the manipulator.At the same time,a simulation and control platform was built for verification.The experimental result shows that the accuracy of 3D modeling is maintained within 8mm,the success rate of obstacle avoidance is 96.5%,and the average planning time is 1.2s,which verifies the feasibility of this method.

关 键 词：工业机器人 视觉定位 三维重建 避障规划 ROS 

分 类 号：TH16[机械工程—机械制造及自动化] TH164[自动化与计算机技术—检测技术与自动化装置] TP242[自动化与计算机技术—控制科学与工程]