Research On 3D Lidar And IMU Tightly Coupled Positioning Algorithm In Static And Dynamic Scenes

With the rapid development of positioning technology,high-precision positioning plays a more and more important role in automatic driving technologyThe role of.In recent years,more and more scholars have invested in the research of location algorithm,resulting in a series of mature industrial products,such as automatic calibration,valet parking,virtual reality and augmented reality.The localization algorithms based on various sensors behave differently in the actual scene,which is determined by the unique nature of the sensor itself.For example,vision based positioning algorithms are easily affected by the lighting conditions of the environment,3D lidar based algorithms will be affected by the number of geometric features in the environment,IMU based algorithms will be affected by their own random walk and noise for a long time,and the positioning accuracy will be worse and worse.Wheel speed based algorithms will have poor robustness of calculation results due to changes in tire pressure and tire slip during cornering;In the fast dynamic scene,the requirements for positioning algorithm are quite high,and the characteristics of different sensors need to be used to learn from each other;In this paper,the combination of 3D lidar and IMU sensor will be selected to build the positioning algorithm.3D lidar is not affected by lighting conditions,and can obtain the original point cloud information of the environment with very high accuracy.IMU sensor is not affected by the change of external environment,and can provide short-time relative pose relationship with high frequency,which will improve the real-time and robustness of the positioning algorithm.In this paper,a 3D lidar and IMU tightly coupled positioning algorithm is built for common autopilot scenes,and an online dynamic point elimination algorithm is proposed to realize high-precision positioning in real time on CPU.In order to further strengthen the robustness,the motion school test module of posture is added.Finally,the algorithm is verified in the real scene.The research contents of this paper are as follows:In order to improve the positioning accuracy and real-time calculation of dynamic scene,an online dynamic point is proposed in this paper.The elimination algorithm adopts multi-resolution depth map to realize the dynamic point elimination from coarse to fine,so as to assist the point cloud registration of 3D lidar odometer,and then make the positioning algorithm obtain higher positioning accuracy.In order to improve the robustness of 3D lidar and IMU tight coupling positioning algorithm,the kinematic consistency test of motion pose is proposed.The calculated pose is modeled to detect the drift degree of odometer.In order to improve the accuracy of 3D lidar and IMU tightly coupled positioning algorithm,a ground feature point extraction method based on normal filtering is proposed,which increases the semantic information of ground feature points and uses three semantic feature points to make the positioning algorithm run reliably.