露天煤矿环境下基于LiDAR/IMU的紧耦合SLAM算法研究

Study on tightly coupled LiDAR-Inertial SLAM for open pit coal mine environment

随着人工智能和无人驾驶等相关学科的快速发展,煤矿装备的智能化和无人化成为了新的趋势。智能设备的应用将大幅提高煤矿作业的生产力以及人员安全性。露天煤矿地形复杂,与城市环境相比无明显的几何特征,具有分段相似性,利用现有以激光雷达为主的同时定位与建图(Simultaneous Localization and Mapping,SLAM)方案在该环境下易出现定位漂移和建图误差较大等现象。针对上述问题,提出了一种基于激光雷达(Light Detection and Ranging,LiDAR)和惯导(Inertial Measurement Unit,IMU)紧耦合的SLAM算法,该算法使用LiDAR和IMU两种传感器作为数据输入,对数据进行预处理,前端利用迭代扩展卡尔曼滤波器将预处理后的LiDAR特征点与IMU数据相融合,并使用后向传播来矫正雷达运动畸变,后端利用雷达相对位姿因子将LiDAR帧间配准结果作为约束因子与回环因子共同完成全局因子图优化。利用开源数据集和露天煤矿实地数据集验证了算法的鲁棒性和精确性。试验结果表明在城市结构化环境中文中所提算法与当前激光SLAM算法精度保持一致,而针对长达两千多米的露天煤矿实地环境,所提算法较FAST-LIO2、LIO-SAM紧耦合算法在定位精度上分别提高了46.00%和23.15%,且具有更高的鲁棒性。

With the rapid development of artificial intelligence and unmanned and other related disciplines,the intelligence and unmanned of coal mining equipment has become a new trend.The application of intelligent equipment will greatly improve the productivity of coal mine operations as well as personnel safety.In this environment,the existing LIDAR-based Simultaneous localization and mapping(SLAM)solution is prone to positioning drift and large mapping errors.To address these problems,a tightly coupled SLAM algorithm based on LiDAR(Light Detection and Ranging)and IMU(Inertial Measurement Unit)is proposed,which uses both LiDAR and IMU sensors as data inputs.The front-end uses an iterative extended Kalman filter to fuse the pre-processed LiDAR feature points with the IMU data and uses backward propagation to correct the radar motion distortion,the back-end uses the LiDAR relative positional factor to use the LiDAR inter-frame alignment results as a constraint factor together with the loopback factor to complete the global factor map optimization.The robustness and accuracy of the algorithm are verified using open source dataset and open pit coal mine field dataset.The experimental results show that the accuracy of the proposed algorithm is consistent with the current LiDAR SLAM algorithm in the urban structured environment,while the proposed algorithm improves the localization accuracy by 46.00%and 23.15%with higher robustness than the FAST-LIO2 and LIO-SAM tightly coupled algorithms for the open pit coal mine field environment of more than 2000 meters long,respectively.