Research On Indoor Environment Mapping And Localization Method Based On UWB/IMU/3D-LiDAR

The application of mapping and localization in the indoor environment involves many aspects of people's life and brings certain economic benefits to the society.The purpose of mapping is to build a map of the environment,which is the basis for other applications such as localization and navigation.Mapping in unknown environment is usually realized based on Simultaneous Localization and Mapping(SLAM),which is more efficient than manual Mapping.When the map information of the target area is known,the online location can be realized according to the real-time input sensor data and the existing map information.The two applications of SLAM-based mapping and location based on existing maps are essentially state estimation problems based on sensor data,but they have different manifestations.It is inevitable that the state estimation method using a single sensor has limitations.In this thesis,the unmanned vehicle platform is used in indoor scenes to complete the above two applications with UWB/IMU/ 3D-LIDAR sensors.The specific research contents are as follows:Firstly,according to the characteristics of mapping in unknown environment and localization in known environment,as well as the working characteristics of sensors used in this thesis,an appropriate multi-sensor fusion state estimation framework is selected.In the slam-based mapping process,state estimation at multiple moments needs to be maintained.In this thesis,factor graph optimization is adopted as the basic framework to fuse the above three sensor data.In the stage of localization based on existing maps,IMU data and li DAR data are fused by filtering framework,and UWB data are input as auxiliary information.The above two applications are dependent on laser point cloud data,in view of the laser point cloud data with noise problem,this thesis design a little bit of cloud pretreatment module,organized on the original input laser point cloud and the ground segment and point glowed class processing,then after processing of point cloud data input to the above two tasks in the framework.In the slam-based mapping,the primary goal is to design the front-end odometry module with low drift and estimate the sensor pose information in unknown environment.For only on the precision of the laser radar odometry is relatively low,in this thesis,partial factor graph was used to optimize the way of IMU data fusion and laser radar odometry data,concrete way is to use the integral model of IMU lidar motion constraints between the key frames in the structure of the local factor for the added company integrating factor in advance;The observation constraints of position and pose structure were obtained by laser point cloud registration,and the key frame laser odometry factor was added to the local factor graph.By optimizing the local factor graph above,low-drift Li DAR-IMU odometry information can be obtained,but the front-end odometry will still accumulate errors after a long time of operation.In this thesis,the global factor graph is used to fuse the front-end odometry data at the backend of SLAM,and UWB ranging information is added as observation constraints to suppress pose errors.The map management module finally constructs a more accurate indoor point cloud map according to the localization position and pose and laser point cloud at different times.In this thesis,the above 3D point cloud map generated based on SLAM is further applied to the location research of known environment.Since the map information is known,the form of state estimation is relatively simple,and there is no need to maintain an incremental map as in SLAM.In this thesis,unscented Kalman filter is used as the basic framework,IMU data is also used to construct motion constraints,and observation constraints are constructed through the registration relationship between laser point cloud and map point cloud,and finally the localization pose of the system is output.In this task,UWB location information is also introduced as the prior value of laser point cloud registration to improve the efficiency of high point cloud registration.In this thesis,the Robot Operating System(ROS)environment is used to verify the effectiveness of the proposed SLAM mapping and the localization method based on existing maps in different indoor scenarios.