Research On Key Technologies Of 3D SLAM For Multi-dimensional Motion Platform

Simultaneous Localization and Mapping(SLAM)in Multi-dimensional motion platform refers to that the robots perceive the three-dimensional environment through sensors in an unknown environment to estimate the six-degree-of-freedom and create a map of the environment.It is one of the key technologies for robot localization and navigation without satellites technology.In order to improve the accuracy of map construction and trajectory prediction as well as the robustness of the system,the research is carried out from feature correlation and multi-sensor fusion.In order to improve the accuracy of correlation,a Li DAR-based odometry algorithm based on related features is proposed.The algorithm extends the detection and matching method of the existing algorithms which detect features in local small area.In feature detection module,the spatial relationship of edge and planar feature points is used to generate stable related features,and the residual function is designed for inter-frame correlation.And the experiments have proved that the algorithm improves the accuracy of mapping.In order to improve the accuracy and robustness of loop closure detection,a loop closure detection algorithm which fusing Li DAR and vision information is proposed.The algorithm uses visual information to detect global loop closure by constructing bag-of-words,and uses Li DAR pose information to detect a local loop closure.Finally,it uses a local map to verify the correctness of the loop closure.The experiments have proved that the algorithm improves the accuracy of loop closure detection and the accuracy of the map.In order to improve the robustness of 3D SLAM algorithm using in multi-dimensional motion platforms,a 3D SLAM algorithm that integrates stereo camera vision information,Li DAR point clouds and the information from inertial Measurement Unit fusion is proposed.The algorithm detects laser features and visual features when interframes are correlated,and constructs a unified residual function to estimate pose.Fusing Li DAR and visual information to detect loop closure.And then a factor graph is constructed using multi-sensor information in the back-end optimization module for nonlinear optimization.Finally,in the mapping module,a multi-resolution octree map is constructed and updated,which realizes the LVI-SLAM algorithm by fusing multi sensors.Experiments show that the algorithm improves the accuracy of trajectory prediction and creates a multi-resolution octree map which have good practical value.Based on the above research,in order to realize the goal of a 3D SLAM system for a multidimensional motion platform,a handheld 3D SLAM system is designed and implemented.The system is divided into a handheld multi-dimensional motion hardware platform and a 3D SLAM software system,which realizes the functions of hardware platform construction,remote transmission of multi-source data,data analysis and mapping.