Research On Mapping And Location Algorithm Based On Multi Sensor Fusion

With the extensive use of mobile robots,the demand for map building and positioning ability is higher and higher.The mapping and positioning algorithm of single sensor often can not meet the demand.Therefore,this paper mainly studies the mapping and positioning method of Lidar based on multi-sensor fusion.For multi-sensor fusion of laser mapping and positioning,pure Lidar mapping and positioning is an important foundation.Therefore,this paper first studies the single Lidar map construction technology,and proposes a SLAM algorithm based on feature point matching.Compared with the previous work,the improvement is as follows: data preprocessing.The accuracy of ground segmentation is optimized by combining RANSAC method,and the angle based point cloud clustering algorithm is improved to reduce the excessive segmentation and feature matching.Based on the point line and point surface matching,the weight of matching information is introduced,and a generating function of matching reliability combining the point cloud clustering information and the point cloud surface information is proposed.This paper optimizes the ground mobile robot.According to the motion constraints of the ground mobile robot,by constructing local sliding window optimization,the accuracy of slam and odometer of the ground mobile robot is prompted.Loop detection uses the latest global point cloud descriptor scheme.Finally,the algorithm proposed in this paper is compared with the open source scheme LOAM,and the data set shows that the proposed method has better performance.Then,this paper extends the proposed single laser SLAM algorithm and realizes the multi-sensor fusion mapping algorithm.For single laser,the following improvements are made: 1.Accurate point cloud motion distortion removal and matching pose prediction are realized by IMU;2.IMU laser fusion odometer based on sliding window optimization;3.GNSS information fusion is realized by a sliding window optimizer decoupled from the front-end fusion odometer.Through the real environment of the experimental data set for testing,and then summarize the data.Secondly,this paper studies the positioning algorithm.Firstly,IMU & GNSS Integrated Navigation Based on ESKF filter is studied and implemented,and a fast and accurate global positioning algorithm based on known map is proposed.The global positioning algorithm proposes a method of combining odometer local positioning and global map joint positioning.The accurate and stable positioning based on point cloud map can be achieved on the map constructed by the method proposed in the previous chapter.For the method proposed in this paper,the key parts of the corresponding experiments are carried out in open source datasets and real scenes and devices.