Abstract

Abstract Aiming at the problems of insufficient accuracy and serious computational time-consumption of mainstream laser odometry schemes, this paper proposes an innovative solution, where the front end is based on the idea of iterative Kalman filtering, which realizes the tight coupling of inertial measurement unit (IMU) and LiDAR to obtain high-precision position and attitude estimation. The back-end optimization employs a factor graph approach, incorporating laser odometry, IMU preintegration, and loop closure detection factors. In comparison with state-of-the-art laser SLAM algorithms, the SLAM algorithm proposed in this paper reduces the absolute position error of trajectories by 9.61%, 91.03%, and 94.89% compared to Fast-lio2, Lego-loam, and a-loam, respectively.