摘要
针对计算资源不足的移动机器人,设计并实现了一种基于边缘计算的视觉同步定位与地图构建(VSLAM)系统。在机器人端引入局部地图,实现机器人位姿估计和轨迹实时跟踪的基本功能,并将具有位姿信息的关键帧发送到边缘计算端;边缘计算端完成全局地图构建、回环检测、地图优化和更新机器人端局部地图等功能。对VSLAM算法的线程进行解耦,提高算法的并行性;采用带有位姿信息的关键帧改进重定位功能。实验采用TUM数据集进行验证,实验结果表明,本文所设计的方法具有更好的实时性、地图构建精度和鲁棒性;当网络通信失败时,机器人可以实现独立定位和轨迹跟踪的基本功能;当网络通信恢复后,VSLAM系统可以快速恢复地图构建和轨迹跟踪功能。
A new visual simultaneous localization and mapping(VSLAM)system based on edge computing is conceived and built to address the mobile robot with insufficient computing resources.The local map is used on the robot side to enable posture estimation and real-time tracking of the robot’s trajectory,as well as delivering keyframes with pose information to the edge computing side.The edge computing side completes the functions of global map construction,loop detection,map optimization and updating the local map of the robot side.Decouple the threads of visual SLAM algorithm to improve parallelism;the keyframe with pose information is used to improve the relocation function.The proposed method offers higher real-time performance,map creation accuracy,and robustness,according to the experimental results using TUM data sets.When network communication fails,the robot can perform fundamental positioning and motion trajectory generation on its own.When the network communication is restored,the visual SLAM can quickly restore the functions of map construction and trajectory tracking.
作者
徐斌
杨东勇
XU Bin;YANG Dongyong(College of Information Engineering,Zhejiang University of Technology,Hangzhou 310023)
出处
《高技术通讯》
CAS
2023年第9期1000-1008,共9页
Chinese High Technology Letters
基金
国家自然科学基金(62072406)资助项目。
