摘要
针对现有即时定位与地图构建(SLAM)算法在煤矿井下巷道退化环境中易出现定位漂移甚至失效的问题,提出了一种融合退化因子的煤矿巷道SLAM算法。该算法以扩展卡尔曼滤波器(EKF)为框架,在惯性测量单元(IMU)与激光雷达(LiDAR)数据的基础上,融合编码器信息实现定位与建图。改进了退化因子计算方法,通过线面特征配准计算特征值与退化因子,通过退化因子的大小表征环境退化程度,实现环境退化评估;设计了基于退化因子的置信度融合机制,可在保持高精度定位建图的同时,显著提高系统鲁棒性;通过提高精度较高特征点权重、降低精度较低特征点权重的方式设计残差,提高退化因子表征准确性。实验结果表明:相比紧耦合激光雷达惯性里程计(LIOSAM)、激光雷达里程计与建图(LOAM)等现有算法,融合退化因子的算法对煤矿退化环境的适应能力更强,可稳定完成定位建图任务;该算法在退化环境下的定位误差为1.222 m,相比LIOSAM算法减小了26.506 m,在非退化环境下的均方根误差(RMSE)均值为0.116 m,低于LOAM和LIOSAM算法;该算法在巷道特征退化路段仍能稳定运行。
To address the problem that existing Simultaneous Localization and Mapping(SLAM)algorithms tend to suffer from localization drift or even failure in degenerated environments of underground coal mine roadways,a coal mine roadway SLAM algorithm incorporating degeneration factors was proposed.The algorithm was built on an Extended Kalman Filter(EKF)framework and integrated encoder information with data from an Inertial Measurement Unit(IMU)and LiDAR to achieve localization and mapping.The method for calculating the degeneration factor was improved by obtaining eigenvalues and degeneration factors through line and plane feature registration,where the magnitude of the degeneration factor characterized the degree of environmental degeneration,thereby enabling environmental degeneration assessment.A confidence fusion mechanism based on degeneration factors was designed to maintain high-precision localization and mapping while significantly enhancing system robustness.The residual was designed by increasing the weights of high-precision feature points and reducing those of low-precision ones,thereby improving the accuracy of degeneration factor characterization.Experimental results showed that,compared with existing algorithms such as the tightly coupled LiDAR-Inertial Odometry via Smoothing and Mapping(LIOSAM)and LiDAR Odometry and Mapping(LOAM),the proposed algorithm demonstrated stronger adaptability to degenerated coal mine environments and could stably complete localization and mapping tasks.The positioning error of the proposed algorithm in degenerated environments was 1.222 m,which was reduced by 26.506 m compared with the LIOSAM algorithm.In non-degenerated environments,its Root Mean Square Error(RMSE)averaged 0.116 m,which was lower than those of the LOAM and LIOSAM algorithms.The algorithm could also operate stably in roadway segments where features were degenerated.
作者
李小波
戴弋杰
刘其鑫
高铭阳
张庆华
LI Xiaobo;DAI Yijie;LIU Qixin;GAO Mingyang;ZHANG Qinghua(CCTEG Chongqing Research Institute,Chongqing 400037,China)
出处
《工矿自动化》
北大核心
2025年第9期133-141,共9页
Journal of Mine Automation
基金
国家重点研发计划项目(2022YFB4703600,2024YFC3013805)
应急管理部重点科技计划项目(2024EMST070703)
贵州省科技计划重大专项项目(黔科合重大专项字[2024]029)
中煤科工集团重庆研究院有限公司院自立科研开发项目(2024YBXM52)。
关键词
井下轮式机器人
即时定位与地图构建
巷道退化环境
退化评估机制
置信度融合机制
退化因子
激光雷达里程计与建图
underground wheeled robot
simultaneous localization and mapping
roadway degeneration environment
degeneration assessment mechanism
confidence fusion mechanism
degeneration factor
LiDAR odometry and mapping
