一种机器人的栅格-语义地图构建方法

A grid-semantic fusion map construction method for mobile robot

基于点云插入或分割的语义地图构建方法存在空间复杂度高、计算资源消耗多等问题,针对以上问题,提出一种移动机器人的栅格语义融合地图构建方法。首先,运用激光雷达构建栅格地图作为融合地图的框架;然后,采用YOLO-v3-tiny获取环境语义信息,并将其属性信息及其在3D环境中的笛卡尔坐标映射至栅格地图,形成上层语义地图;最后,通过随机采样和K-means对空间区域进行划分。实验结果表明,所提方法能实时获取环境结构的信息和语义信息,所构建的融合地图为目标搜索任务提供语义信息引导,相比穷举式目标搜索,减少了74.8%的时间代价,提高了目标搜索任务的执行效率。

The semantic map construction method based on point cloud insertion or segmentation has some shortcomings,such as high spatial complexity and high consumption of computing resources.To solve the above problems,this paper proposes a grid semantic fusion map construction method for a mobile robot.Firstly,the grid map constructed by laser radar is used as the basic framework of fusion map;then,the environment semantic information is collected by using the YOLO-v3-tiny algorithm,and its attribute information and Cartesian coordinates in the 3D environment are mapped to the grid map to form the upper semantic map;finally,the spatial region is divided by random sampling and K-means.Experimental results show that the proposed algorithm can extract the environment structure information and semantic information,and the constructed fusion map can provide semantic information guidance for the target search task,which reduces the time cost by 74.8%compared with exhaustive target search,it also improves the execution efficiency of the target search task.