不规则多物品无序分拣机器人研究  

Research on Picking Robot of Various Irregular Objects in Disorder

作　　者：武帅 

机构地区：[1]合肥哈工慧拣智能科技有限公司,安徽合肥230000 [2]哈工大机器人(合肥)国际创新研究院,安徽合肥230000

出　　处：《工业控制计算机》2023年第8期69-71,共3页Industrial Control Computer

摘　　要：针对多种不规则物品混乱摆放的抓取问题,一直是机器人领域研究的热点和难点,对其研究在家庭生活服务机器人等领域有很高的应用价值。为实现这一功能,基于双目相机和协作机械臂搭建一套物品分拣机器人系统。系统中基于YOLOv5算法和PSMNet网络建立起目标检测和双目立体匹配算法,利用ROS系统的MoveIt模块进行机械臂控制。经过实验验证,该机器人能够有效识别、定位和分拣形状不规则、表面纹理丰富、摆放混乱的物品,对经过训练的物品识别成功率达到100%,分拣成功率达到93.3%。Grasping a variety of irregular objects in disorder has always been a hot and difficult problem in the field of robot research,and has high application value in the field of family life service robot.In order to realize this function,a set of objects picking robot system is built based on binocular camera and cooperative robot arm.In the system,object detection and binocular stereo matching algorithms were established based on YOLOv5 and PSMNet algorithm.MoveIt module of ROS system is used to control the manipulator.After experimental verification,the robot can effectively recognize,estimate pose and pick the objects with irregular shape,messy surface texture and disordered arrangement,the success rate of recognition of trained objects reached 100%,the success rate of picking reached 93.3%.

关 键 词：机器人抓取 双目视觉 深度学习 无序分拣 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP242[自动化与计算机技术—计算机科学与技术]