| At present,as mobile robots gradually play an important role in industry,agriculture,military and other fields.The robot’s ability to perceive its environment has become a key factor in determining its level of intelligence.The robot’s perception of the operating environment is the basis for landing point planning,path planning and other tasks.In this paper,the wheeled robots is used as the carrier and apply multiple sensor information fusion to improve the robot localization accuracy and the quality of environment map construction,and then enhance the mobile robot’s ability to perceive the surrounding environment.Firstly,the current state of research on robot localization and mapping is analyzed,and the main contents of this paper are proposed.According to the main content of this paper,the sensor coordinate system and its measurement model are described,and the sensor model and coordinate system definition are taken as the theoretical basis of multi-sensor fusion.Secondly,a Li DAR-Inertial odometry localization system based on Li DAR/IMU fusion is developed for the accurate estimation of robot poses.The IMU data is applied for Li DAR point cloud distortion correction,and the deskewed point cloud is extracted by curvature for feature point classification.The Li DAR point cloud key frames are extracted,the sliding window method is introduced,the residual equation is constructed based on the corner pointline and planar point-plane matching methods,and the Li DAR odometry factor is obtained by using LM optimization;the IMU pre-integration factor is obtained by pre-integrating the IMU data between key frames based on the IMU model;through the Li DAR keyframs loop closure detection,the loop closure factor is obtained.The factor graph optimization of the three factors is performed to obtain the global pose of the robot.Experiments are designed to compare with the Li DAR odometry using a wheeled robot as a platform to demonstrate the accuracy of the constructed Li DAR-inertial odometry localization.Thirdly,the point cloud fusion of Li DAR/depth-camera is realized in ROS environment,and the fusion point cloud is used as the original measurement data when the robot map is constructed to improve the quality of map construction.In the synchronization of Li DAR/depth-camera information: Using the timestamp approximation mechanism to achieve time synchronization;Using the calibration plate as the medium,applying feature point matching to achieve sensor space synchronization.In view of the problems of many noise points and large measurement errors in the camera pointcloud,a variety of filtering and invalid point removal algorithms are used to preprocess the point cloud.Combined with the GICP algorithm which is limited by translation and rotation thresholds,the fusion point cloud is realized.The ROS topic communication is used to realize the fusion point cloud topic output.In the experiment,the two point clouds are matched and fused under different application environments of the robot.Finally,aiming at the problem of 2D map height information loss,the construction principle and construction method of the 2.5D local elevation map and 3D octree map are described.In the local elevation map construction,the fusion point cloud data is used to update the map measurement value;the Li DAR-inertial odometry is used to estimate the robot’s localization to realize the motion update of the map.The octomap reflects the feasibility of the map by probability.The map uses the fused point cloud data as the basis for the presence of obstacles,and realizes the splicing and construction of the global map by the Li DAR-Inertial odometry.In the experimental part,among the two types of maps described,the wheeled robot is equipped with multiple sensors as a platform,the Li DAR-Inertial odometry is used to estimate the position of the robot and the Li DAR point cloud,camera point cloud and fusion point cloud are used as map measurements to compare the quality of map construction,proving that the method used in this paper effectively improves the robot’s environmental perception ability.Finally,the experimental results are presented centrally through the Qt interface to improve the integration of the system. |
