IMU/ODO Pre-integration Based Tightly Coupled LiDAR SLAM in Degradation Scenarios

Aiming at the problems of localization failure, overlapping drift of map construction and inability to run in real time in degraded scenes (e.g., long straight corridors and tunnels) lacking environmental feature textures, this paper proposes a tightly-coupled laser SLAM algorithm based on Inertial Measurement Unit/wheel speed odometer (IMU/ODO) pre-integration. The method extracts line and surface feature points based on the removal of point cloud distortion by IMU/ODO pre-integration. Based on the line and surface constraints, the point cloud alignment residual function is constructed, and the Gauss-Newton method is used to optimize the residuals to obtain the carrier motion estimation. Meanwhile, the degradation direction of the point cloud is estimated based on the perturbation model, the degradation degree is quantitatively analyzed, and the fusion weights are adjusted in real time according to the degradation degree. Combining IMU/ODO pre-integration and Scan-Context loopback detection, the factor graph is constructed for global optimization. Vehicle-mounted experiments show that the algorithm in this paper effectively overcomes the problems of low positioning accuracy of SLAM algorithm and overlapping drift of map construction under degraded environment, and the algorithm is able to output accurate position in degraded environment, and the overall error is less than 4% in 28 m long straight corridor. The algorithm is able to build clear and accurate maps in the map building experiment in the back-type corridor environment.