Research On GNSS/Lidar/IMU Integrated Vehicle Positioning Technology In Complex Urban Environment

Simultaneous localization and mapping(SLAM),that is,estimating the position and att itude of the carrier and constructing the environmental point cloud map without a priori bas ed on the lidar sensor data,has the advantages of high precision,not affected by light and su itable for indoor and outdoor scenes.It is a research hotspot in the field of unmanned syste m positioning at present.This paper studies GNSS/lidar/IMU fusion positioning technology for the demand of high-precision positioning of intelligent vehicles.The position and attitud e estimation algorithm of vehicle borne laser radar is improved according to the characterist ics of different cloud components;The pre integration strategy is used to add IMU observati on data to improve the robustness of laser pose estimation algorithm in outdoor open scene;Global navigation satellite system(GNSS)is integrated through graph optimization algorith m,so that it can maintain positioning continuity for a long time in complex urban environm ent and eliminate the cumulative error of laser slam system.The specific research contents a nd contributions of this paper are as follows:(1)An improved step-by-step pose estimation method based on vehicle platform is pro posed.By deriving the calculation formula of laser slam pose,aiming at the difference betw een ground feature points and edge feature points in solving the pose of each degree of freed om,two kinds of feature points are used to solve the pose of the vehicle respectively.Firstly,the ground feature points are used to solve the three degrees of freedom pose,and then the edge feature points are used to solve the other three degrees of freedom pose,and the first st ep pose solution results are optimized to realize the optimization of lidar pose estimation alg orithm for vehicle platform.In the data set and actual environment test,this scheme improv es the positioning accuracy by more than 7% and up to 45% on the premise of ensuring that the time-consuming is better than the original algorithm.(2)The laser inertial pose estimation algorithm based on inertial measurement unit(IM U)preintegration is studied.Aiming at the problems of low accuracy and poor robustness of laser slam in high dynamic and open scenes,the laser inertial pose estimation algorithm bas ed on tight coupling is studied.Through IMU preintegration algorithm,the carrier motion st ate and IMU parameters are optimized in the fusion process to improve the positioning accu racy of lidar in high dynamic and open scenes.Experiments show that the positioning accur acy of laser inertial pose estimation algorithm is more than 25% higher than that of lidar pos e estimation algorithm,and the robustness of the system is greatly improved.(3)A GNSS/lidar/IMU multi-source sensor fusion framework based on Ceres graph op timization is constructed.In this paper,the interpolation of laser inertial pose estimation res ults is proposed to improve the utilization of satellite positioning system observation data;T he Jacobian matrix of the difference of pose increment and relative pose component of diffe rent systems is derived,and the optimal solution of pose is solved by graph optimization str ategy;The sliding window algorithm is used to update the estimated value of the conversion matrix between systems,so that the current system pose estimation results accord with the global optimization.The algorithm proposed in this paper aims at the problem that the error of inertial navigation position and attitude estimation for a long time diverges too fast when the satellite signal of GNSS/IMU integrated positioning system is missing.A scheme of inte grating inertial element and lidar observation is proposed to suppress the error divergence of inertial element and improve the continuity and reliability of positioning system;GNSS sys tem is used to eliminate the cumulative error of laser inertial pose estimation algorithm,and the positioning results are unified into the navigation coordinate system.The test results sho w that the pose estimation accuracy of the proposed algorithm is more than 40% higher than that of the laser inertial pose estimation algorithm,and reaches the accuracy of meter level;Compared with the factor map fusion algorithm based on gtsam,the positioning accuracy of the algorithm proposed in this chapter is significantly improved compared with satellite and laser inertial positioning algorithms in complex urban environment.