“小巧手”自动收纳机器人抓取分析及评测

Grabbing Analysis and Evaluation of“Smart Hand”Automatic Storage Robot

“小巧手”自动收纳机器人搭建在Jetson TX2主板的嵌入式平台下,在传统的机械手中安装双目摄像机,通过视觉系统架构设计实现目标抓取。双目摄像机对空间布局进行扫描,视觉检测系统平台和线结构光扫描测量平台在像素坐标系、图像坐标系、摄像机坐标系、世界坐标系等特有坐标系下完成路径规划。立足卷积神经网络(CNN)进行视觉分类,基于SSD模型进行训练、校验和部署,实现CNN框架下的物体识别。开展系统电路及机构设计,机构底盘搭载了防跌落、红外测距等多种传感器,通过爪力感知系统对抓取物体操作进行抓取力的调节,最后将物体放置于指定地点完成收纳。

The“smart hand”automatic storage robot was built on the embedded platform of Jetson TX2motherboard,and a binocular camera was installed in the traditional robot hand,so that the target could be captured by the architectural design of the visual system.The binocular camera scanned the spatial layout,and the vision inspection system platform and the line structured light scanning measurement platform completed the path planning in the specified coordinate systems,i.e.the pixel coordinate system,the image coordinate system,the camera coordinate system and the world coordinate system.Convolutional neural network(CNN)was used for visual classification,SSD model was used for training,verification and deployment,and the object recognition under CNN framework was realized.The designs of system circuit and mechanism were carried out.The chassis of the mechanism was equipped with various sensors such as anti-falling,infrared ranging,etc.Through the claw force sensing system,the grasping force of grasping objects was adjusted.Finally,the objects were placed in the designated place to complete the storage work.