B-Spline-Based Pose Estimation And Map Construction With Tight Coupling Of LiDAR And IMU

The Li DAR and Inertial Measurement Unit(IMU)are two types of sensors often equipped with intelligent unmanned systems.Lidar can provide high-precision ranging information,but its low data output frequency cannot meet the need for reliable positioning in the fast motion state of the carrier.Although the IMU can measure the acceleration and angular velocity of the output line at a higher frequency and continuously estimate the carrier pose accordingly,it can only accurately estimate the posture for a short period of time as an internal sensor.In order to achieve the efficient fusion and complementary advantages of lidar and IMU,this thesis proposes a B-spline-based laser-inertial data tightly coupled pose estimation and map construction method,which can effectively improve the accuracy and effectiveness of the Simultaneous Localization and Mapping(SLAM)algorithm.The laser SLAM algorithm implemented in this thesis includes two modules,odometer and graph,of which the odometer module is based on the open source SLAM solution LIOmapping implementation,and the IMU pre-integration and laser feature point matching pair are jointly optimized to initially solve the motion posture through sliding window technology.In order to introduce the IMU high-frequency measurement constraint into the mapping module,this thesis proposes a laser-inertia data tight coupling optimization method based on the B-spline motion trajectory expression in the mapping module to increase the connection between the positions and postures of adjacent lidar to be optimized,and takes the posture estimation result of the odometer module as the initial value of optimization to the final highprecision pose.Trajectory optimization problems based on B-splines have more complex derivative expressions,and solving by using the automatic derivative method of the Ceres library can avoid giving derivative expressions,but its calculation efficiency is poor and cannot meet the real-time performance requirements.Therefore,this thesis also derives the display expression of the laser residual derivative in the optimization solution process based on B-spline,and cooperates with the Ceres library to achieve an efficient optimization solution process.In this thesis,the proposed method is tested based on the indoor and outdoor environment datasets and the Oxford public dataset collected by the self-developed equipment,and the experimental results verify that the trajectory and mapping accuracy of the proposed method are better than the existing laser-inertial tight coupling method LIO-mapping,and the optimization efficiency of the analytical and derivative method implemented in this thesis is higher than that of the traditional automatic derivative method.