Research On Positioning And Mapping Method Of Patrol Robot Based On 3D Lidar

With the continuous development of artificial intelligence and robot technology,some environmental inspection work such as campus,substation,coal mine,cable has been replaced by robots.In the actual working process of inspection robots,it is difficult to estimate the complexity of the surrounding structured and unstructured environments in advance,thus increasing the difficulty in achieving high-precision positioning and mapping.Aiming at the above problems,this thesis conducts an in-depth study on the positioning and mapping methods of inspection robots based on three-dimensional lidar.The specific contents are shown as follows:(1)Aiming at the problems of low accuracy and easy drift of laser odometer constructed by traditional three-dimensional point cloud registration algorithm in complex environment,an adaptive laser odometer for complex environment is designed in this thesis.Firstly,after processing the original point cloud data collected by 3D laser radar through the point cloud preprocessing link,the ground segmentation method is used to complete the point cloud data segmentation and obtain the richness information of the road surface point cloud.Then,the NDT algorithm is used to minimize the distance between the front and back frame point cloud data,and the coarse registration of point cloud data is realized.Finally,an appropriate ICP algorithm is selected according to the surrounding environment to complete the high-precision registration of point cloud data,and an adaptive laser odometer is constructed by using the point cloud transformation relationship output from the registration link.(2)In order to make the inspection robot obtain more accurate and efficient global positioning,this thesis proposes a multi-sensor combination positioning and mapping method based on graph optimization.Firstly,the constraints of the output position and pose of the adaptive laser odometer,the pre-integral constraints of the inertial measurement unit and the constraints of loopback detection using the Scan Context algorithm were integrated into the graph optimization model.Then,the optimal pose estimation is obtained through g2 o library back-end solution.Finally,an accurate 3D point cloud map is constructed according to the estimated results and real-time location is realized by using the sliding window method based on marginalization.(3)In order to verify the feasibility and effectiveness of using the above methods for positioning and mapping of inspection robots in complex environments,wheeled mobile robots were set up in this thesis as the experimental platform.The human-computer interaction software developed by ros-qtc-pluging plug-in of Qtcreator is used to remotely control the inspection robot to collect point cloud data in structured and unstructured environments respectively.Through a large number of real vehicle tests in different environments,it can be concluded that the average displacement error of this method is 0.021 m in the structured environment and 0.1m in the unstructured environment.