关键词: 激光雷达；因子图优化；IMU；紧耦合；SLAM 算法

Abstract: In recent years, the field of unmanned driving has become a hot topic of widespread concern, and the same time, simultaneous localization and mapping (SLAM) technology is the basis of high-precision map creation and unmanned vehicle navigation, and the current mainstream laser SLAM algorithm can basically meet the needs of applications, but there is still a drift problem in the process of large-scale scene mapping, and there is still room for improvement in algorithm lightweight and real-time. In this paper, the simultaneous localization and mapping algorithm of lidar and inertial measurement unit (IMU) tightly coupled is mainly studied, and the front-end part mainly removes distortion and feature extraction of laser point cloud data, and the back-end uses factor map to fuse IMU pre-integration factor, laser odometry factor and loopback detection factor for fusion pose output. In order to improve the real-time performance of the algorithm, this paper uses the iKD-Tree data structure to maintain a local map, and uses the Fast-generalized iterative closest point (GICP) algorithm to solve the loopback detection pose transformation. The test of Kitti's public dataset shows that the proposed algorithm improves the real-time and robustness of the algorithm while ensuring accuracy.

Key words: Lidar; Factor graph optimization; IMU; Tight coupled; SLAM algorithm