摘要
针对单机器人同时定位与地图构建(Simultaneous Localization And Mapping,SLAM)算法构图效率低的问题,设计了基于ORB-SLAM3的多机器人协作SLAM的实时融合方案。首先,向两个SLAM进程输入相同的关键帧并以此帧所在位姿为世界坐标系完成初始化,然后各机器人以此坐标系为基础完成定位于建图,同时将地图汇总融合完成任务。此方案避开了传统地图融合问题中特征匹配与位姿估计的技术难点。在Gazebo仿真环境进行测试实验,验证了方案的有效性、实时性和鲁棒性。
To solve the problem of single robot's inefficiency in Simultaneous Location And Mapping(SLAM),this paper designs a solution to solve real-time map merging of multi-robot cooperative SLAM based on ORB-SLAM3.First,enter one key frame into the two SLAM processes,and the posture of this frame is used as the world coordinate system to complete initialization.Then,each robot completes the mapping based on this coordinate system,and makes the maps merged at the same time.This solution avoids the technical difficulties of matching features and estimating posture in the problem of traditional map merging.The effectiveness,real-time performance and robustness of the solution are verified by testing in the Gazebo simulation environment.
出处
《工业控制计算机》
2021年第6期64-65,67,共3页
Industrial Control Computer
基金
广东工业大学大学生创新训练项目-多机器人协同SLAM研究。
