机器人远程点云分割框架设计与实现

Design and Implementation of Robot Remote Point Cloud Segmentation Framework

针对机器人端运算能力有限、难以快速进行3D点云分割的问题,提出了一种基于远程过程调用(remote procedure call,RPC)的点云远程分割框架,实现了机器人端快速获取点云分割结果的功能。首先位于机器人端采集点云数据;然后把点云序列化为Protocol Buffers的格式;服务端对点云采样、分割并把结果返回至机器人端;最后机器人端获取服务端结果并反序列化得到点云分割结果。实验表明,该方法和常用的RESTful接口相比,在保障实时性的基础上,能大幅度减少流量消耗,从而提升了机器人点云分割的作业效率。

To solve the problem that the robot has limited computing resource and the difficulty to quickly perform 3D point cloud segmentation,a remote point cloud segmentation framework based on remote procedure call(RPC)is proposed to realize the function of quickly obtaining the point cloud segmentation result on the robot side.First,the point cloud data is collected on the robot side;then the point cloud is sampled and serialized to the Protocol Buffers format and sent to the server for segmentation;finally,the robot side deserializes the point cloud segmentation results from server.Experiments show that compared with normally used RESTful interfaces,this method can greatly reduce resource consumption and improve the efficiency of robot point cloud segmentation on the basis of ensuring real-time performance.