Research On Simultaneous Localization And Mapping Of Outdoor Inspection Robot Based On Multi-sensor Fusion

In recent years,with the continuous development of mobile robots,outdoor inspection robots have gradually been used in daily life.Traditional outdoor mobile robots are positioned based on Global Positioning System(GPS)or Inertial Navigation System(INS).However,when GPS signals are affected by factors such as non-lineof-sight and multipath,traditional There is a large deviation in the GPS positioning method.In addition,when the mobile robot travels into a GPS-denied environment,it will be impossible to obtain GPS positioning data.Therefore,people are gradually using Simultaneous Localization and Mapping(SLAM)technology based on visual cameras or lidar to locate and navigate outdoor inspection robots.Compared with the shortcomings of visual cameras that are easily affected by changes in light intensity,lidar is not sensitive to light intensity and can accurately perceive the surrounding environment in different outdoor environments,fundamentally ensuring the accuracy of the laser SLAM system in outdoor environments.The SLAM system with lidar as the core has become the solution of most outdoor mobile robot companies.In addition,since the SLAM system using only lidar cannot solve the problems of environmental degradation and low sampling frequency,an Inertial Measurement unit(IMU)is often used to solve this problem in small and medium-scale scenarios.However,the measurement error of the inertial measurement unit will increase with time,and the accuracy of the SLAM system using lidar and inertial measurement unit will decrease in a large-scale scene.In order to improve the accuracy of the laser inertial SLAM system,the method of fusing GPS data is often used to eliminate the cumulative error.This dissertation studies the problem of the accuracy degradation of the multisensor fusion SLAM algorithm with lidar as the core in the case of GPS interruption in the outdoor environment of the science and technology park for inspection robots.Taking lidar as the core,a multi-sensor fusion SLAM method combining lidar,IMU and GPS is proposed.First,in order to build a complete multi-sensor fusion algorithm,a data sampling model and coordinate system are established for the inspection robot chassis and each sensor.At the same time,the external parameters of the coordinate system between the lidar,IMU and GPS are calibrated,and the internal parameters of the IMU are calibrated,so as to more accurately fuse the sensor measurement data of the lidar,IMU and GPS.Finally,the method of combining software and hardware is used to realize the time synchronization of lidar,IMU and GPS.Secondly,aiming at the problem that the front-end odometer based on the traditional laser SLAM system has poor accuracy during long-term operation,and in order to make the robot’s positioning more accurate when the inspection robot travels to a GPS-denied environment and the GPS data is interrupted,a method using Iterative Error Kalman Filter(IESKF)tightly couples the front-end odometry of lidar and inertial measurement unit to improve the accuracy of robot pose estimation using only lidar and inertial measurement unit in outdoor scenes.And experimental evaluation with traditional front-end odometry on KITTI dataset.It is proved that the accuracy of IESKF front-end odometry in outdoor scenes is better than that of traditional front-end odometry.Then,there is still room for improvement in the global accuracy of the laser inertial odometry based on IESKF,and in order to further improve the positioning accuracy and robustness of the algorithm,a method of integrating laser radar,inertial measurement unit and GPS using factor graph optimization is proposed.In the factor graph optimized SLAM system,in order to reduce the calculation load generated by graph optimization,the key frame and sliding window strategy is adopted,and the laser inertial odometry factor and IMU pre-integration factor obtained by IESKF are added to the factor graph,and in order to further Reduce the cumulative error,add loopback detection factor and GPS factor,and screen GPS data based on GPS status and confidence at the same time when GPS data fluctuates to prevent adding abnormal GPS data.Finally,in order to verify the accuracy and robustness of the multi-sensor fusion SLAM system proposed in this dissertation,the open source data set KITTI and the self-built inspection robot platform are tested in two cases of GPS uninterrupted and GPS interrupted,and compared with other The open source SLAM algorithm evaluates trajectory and pose.The results show that the SLAM system proposed in this dissertation has better accuracy in both cases of GPS uninterrupted and GPS interrupted.