基于改进EKF_LOAM的电缆沟巡检机器人精准定位策略  

The precise positioning strategy of cable trench inspection robot based on improved EKF_LOAM

作　　者：双丰 马翰林 杨杰 李少东 SHUANG Feng;MA Hanlin;YANG Jie;LI Shaodong(Guangxi Key Laboratory of Intelligent Control and Maintenance of Power Equipment,GuangxiUniversity,Nanning 530004,China;Nanning Power Supply Bureau,Guangxi Power Grid,Nanning 530021,China)

机构地区：[1]广西大学广西电力装备智能控制与运维重点实验室,南宁530004 [2]广西电网有限责任公司南宁供电局,南宁530021

出　　处：《中国惯性技术学报》2024年第4期326-335,共10页Journal of Chinese Inertial Technology

基　　金：广西自然科学基金—青年基金(2023GXNSFBA026069);广西高校中青年教师科研基础能力提升资助项目(2022KY0008)。

摘　　要：精确定位和准确构建地图(SLAM)是电缆沟巡检机器人顺利完成巡检任务的基础。传统激光SLAM算法在电缆沟环境中容易产生“长廊效应”,且容易在z轴方向上产生漂移。针对上述问题,提出了一种基于多传感器融合的改进扩展卡尔曼滤波激光雷达测距与制图(EKF_LOAM)算法。通过扩展卡尔曼滤波器(EKF)将车轮里程计和惯性里程计融入LeGO_LOAM算法框架中来抑制“长廊效应”,在EKF_LOAM算法的基础上引入IMU的z轴线速度测量值,并设计基于特征点数量的自适应协方差计算方程,约束由面特征点严重缺失引起的z轴漂移。分别在虚拟和真实电缆沟环境中开展实验,结果显示所提算法的最终行进误差比LeGO_LOAM算法的行进误差减小了超过40%,比Cartographer算法的行进误差减小4.57%,z轴方向误差比EKF_LOAM算法减小49%,证明了所提算法在电缆沟环境中优于传统的LeGO_LOAM、Cartographer和EKF_LOAM算法,更适用于电缆沟巡检任务。Accurate positioning and precise map construction are fundamental for the successful completion of inspection tasks by cable trench inspection robots.Traditional laser SLAM algorithms often encounter challenges such as the'long corridor effect'and z-axis drift in cable trench environments.To address these issues,an improved EKF_LOAM algorithm,utilizing multi-sensor fusion,is proposed.By extending the Kalman filter(EKF)to integrate wheel odometry and inertial odometry into the LeGO_LOAM framework,the'long corridor effect'is suppressed.Additionally,based on the EKF_LOAM algorithm,the Z-axis velocity measurement of IMU is introduced,and the adaptive covariance equation based on the number of feature points is designed to constrain the Z-axis drift caused by the serious absence of surface feature points.Based on the EKF_LOAM algorithm,the IMU measurements of z-axis linear velocity is introduced,and the adaptive covariance equation based on the number of feature points is designed to constrain the z-axis drift caused by significant feature point loss.Experimental results conducted in both simulated and real cable trench environments demonstrate that the proposed algorithm reduces the final positioning error by over 40%compared to the LeGO_LOAM algorithm and by 4.57%compared to the Cartographer algorithm.Furthermore,the z-axis direction error is reduced by 49%compared to the EKF_LOAM algorithm,indicating the superiority of the proposed algorithm in cable trench environments over traditional LeGO_LOAM,Cartographer,and EKF_LOAM algorithms,making it more suitable for cable trench inspection tasks.

关 键 词：电缆沟巡检 EKF_LOAM 激光SLAM 长廊效应 z轴漂移 

分 类 号：TP242.3[自动化与计算机技术—检测技术与自动化装置]