Research On High Precision Mapping And Positioning For Unmanned Engineering Vehicle

The intelligent,unmanned and collaborative development of engineering vehicles is the direction of industrial development,and the construction and positioning of highprecision maps for non-structured environments is the key to achieving these development goals.The complex surface topography,large area and harsh working conditions of unstructured environments pose many challenges to map construction and positioning:the vehicle-mounted LiDAR is difficult to align under terrain excitation,the semantic segmentation method under structured roads is not fully applicable to unstructured environments,the GPS signal is susceptible to interference,and the IMU odometer accumulates large errors under strong random vibration,etc.To address the above problems,this thesis realizes unstructured terrain construction based on UAV tilt photography technology,realizes terrain semantic segmentation based on digital gradient model and vehicle-terrain contact coupling relationship,and proposes a coupled UWB quadrilateral measurement and map matching positioning method.Ultimately,high-precision map construction and high-precision positioning in unstructured environments are achieved,providing a navigation basis for autonomous driving and autonomous operation of engineering vehicles.The specific work of the thesis is as follows.(1)Unstructured terrain modelling based on UAV tilt photography.The operating sites of engineering vehicle change dynamically as construction progresses.In order to realise the problem of rapid and highly accurate modelling of non-structural environments,the thesis achieves rapid construction of non-structural terrain by means of UAV tilt photography.Further,the digital elevation model and the 2D RGB model of the unstructured terrain are constructed by filtering out outliers and invalid weed point clouds through post-processing processes such as statistical filtering in geometry and colour space,and area growth hole completion(2)Terrain semantic information extraction for the driving stability of unmanned engineering vehicles.For semantic information extraction under unstructured terrain,this thesis achieves semantic extraction of terrain obstacles based on the maximum between-class variance of the digital gradient model.Further,the extended steady-state margin angle is proposed and the steady-state metrics are solved based on the vehicleterrain contact coupling relationship to classify the regional safety semantic information according to the vehicle steady-state.Finally,a multi-information layer high-precision map construction containing terrain model and semantic information is realized in the unstructured environment.(3)Coupled positioning based on UWB quadrilateral measurements and map matching.To address the problem of high-precision localization in unstructured environments,this thesis proposes a Coupled positioning method based on UWB quadrilateral measurements and map matching.Primary positioning of unmanned engineering vehicles by UWB quadrilateral measurement.Under the neighbourhood constraint of primary positioning,secondary map matching positioning is achieved,and the base station ranging error is calibrated based on the secondary positioning data to improve the primary positioning accuracy,forming a loopback coupling to achieve high accuracy positioning in non-structural environments.In summary,this thesis focuses on the construction of high precision maps and localisation for unmanned engineering vehicles in unstructured environments,and verifies the effectiveness of this method through experiments.