摘要
煤矿智能化是煤炭行业高质量发展的技术支撑,关键岗位的机器人替代是实现煤炭少人化、无人化的高效开采的发展趋势。即时定位与地图构建(Simultaneous Localization and Mapping,SLAM)是煤矿机器人自主移动与导航的关键技术之一。煤矿井下为典型非结构化环境,空间狭长局促,结构复杂多变,照明情况不均匀,对煤矿井下SLAM提出了严峻挑战。总结了煤矿井下地图构建研究现状,针对LeGO-LOAM算法的回环检测仍存在的不足,利用SegMatch算法改进LeGO-LOAM的回环检测模块,且使用ICP算法进行全局图优化,提出了一种融合LeGO-LOAM和SegMatch的改进算法,阐述了该算法的原理和实现步骤;开展了煤矿井下模拟场景试验,对比分析改进前后SLAM算法的建图效果以及精度,试验结果表明改进算法构建的地图回环效果更好,估计轨迹更平滑、精确;结合导航需求研究了二维占据栅格地图的构建方法,试验验证了该方法所构建的栅格地图精度,结果表明有效滤除动态障碍物等离群噪点后的栅格地图具有0.01 m的建图精度,且所需存储空间较点云地图降低了3个数量级。研究成果有助于煤矿井下非结构环境下SLAM和煤矿机器人实时定位和自主移动。
Intellectualization of coal mine is the technical support for high-quality development of coal industry,and robot replacement of key posts is the development trend of realizing efficient mining of coal with few people and no people.Simultaneous localization and mapping(SLAM)is one of the key technologies for autonomous movement and navigation of coal mine robots.The environment of underground coal mine is a typical unstructured environment,with narrow space,complex and changeable structure and uneven lighting,posing a severe challenge to the realization of SLAM in the underground coal mine.The research status of the map construction of the underground coal mine is summarized.In view of the shortcomings of the loopback detection of the LeGO-LOAM algorithm,the SegMatch algorithm is used to improve the loopback detection module of the LeGO-LOAM,the ICP algorithm is used to optimize the global map,and an improved algorithm integrating LeGO-LOAM and SegMatch is proposed,and the principle and implementation of the algorithm are discussed.The underground simulation scene experiments of coal mine were carried out,the mapping effect and accuracy of SLAM algorithm before and after the improvement were compared and analyzed,and the results showed that the map loopback effect constructed by the improved algorithm was better,and the estimated trajectory was smoother and more accurate.The construction method of two-dimensional occupied grid map is studied aiming at the navigation requirements,and the accuracy of the grid map constructed by this method is verified through experiments.The results show that the grid map after effectively filtering outliers such as dynamic obstacles has a mapping accuracy of 0.01 m,and the required storage space is 3 orders of magnitude lower than that of the point cloud map.The research results are helpful to the realization of SLAM and real-time positioning and autonomous navigation of the coal mine robot under the unstructured environment of the underground coal mine.
作者
薛光辉
李瑞雪
张钲昊
刘爽
魏金波
XUE Guanghui;LI Ruixue;ZHANG Zhenghao;LIU Shuang;WEI Jinbo(School of Mechanical Electronic and Information Engineering,China University of Mining and Technology,Beijing 100083,China;Key Laboratory of Intelligent Mining and Robotics,Ministry of Emergency Management,Beijing 100083,China;Institute of Intelligent Mine and Robotics,China University of Mining and Technology,Beijing 100083,China)
出处
《煤炭科学技术》
EI
CAS
CSCD
北大核心
2023年第8期219-227,共9页
Coal Science and Technology
基金
国家自然科学基金资助项目(51874308,61673385)。