Estimation And Prediction Of Position And Attitude Of Engineering Vehicles Based On Priori Information Of Global Terrain Environment

Engineering vehicles play a crucial role in the infrastructure construction due to their high load-bearing capacity,high efficiency and adaptability to the environment.In recent years,as society continues to develop and technology continues to advance,engineering vehicles are also developing in the direction of intelligence,unmanned and collaborative,in order to achieve a safe,energy-saving,efficient and fully autonomous driving operation mode.In unknown environments,engineering vehicles rely on the accurate estimation of the vehicle’s position and attitude at the current moment as well as at future moments for rational path planning,speed optimization and autonomous and safe obstacle crossing as the crucial key to their unmanned and intelligent implementation.However,engineering vehicles often drive in unstructured terrain with complex topography,and the non-linear variation of position and attitude parameters during the driving process is obvious.In addition,the driving speed of engineering vehicles is low,the signal-to-noise ratio is not high and there are static bias problems,coupled with the high measurement cost and the difficulty of direct measurement of state parameters,all these factors bring certain difficulties and challenges to the accurate estimation and prediction of position and attitude state parameters during the driving process of engineering vehicles.This dissertation focuses on the position and attitude estimation and prediction of engineering vehicles in unstructured terrain environment,constructs a 3D global unstructured terrain environment model,and uses this model as a priori information to match with the local terrain environment model constructed in real time to achieve the position and attitude estimation of engineering vehicles,on this basis,further proposes a method for the position and attitude prediction of engineering vehicles,which provides an important theoretical and practical basis for the realization of autonomous driving and intelligent operation of engineering vehicles.The specific works are as follows.Rapid global map construction for unstructured terrain environments.Rapid global modelling of unstructured terrain environments through methods such as UAV tilt photography techniques and feature fusion algorithms.It is designed to address the problems of matching failures caused by image affine distortion and perspective aberration due to large angular photographic tilt angles in large-scale terrain modelling,as well as the variability of local features and lack of structured features in unstructured terrain environments.The Multi-Scale Retinex(MSR)algorithm is first introduced to improve the color recovery factor for feature enhancement of the image.Maximally Stable Color Regions(MSCR)are then obtained by fitting regions using the image moment method and normalized.Then combined with an optimized Speed Up Robust Feature(SURF)to achieve fast and accurate feature extraction and matching between the affine images.Finally,the global unstructured terrain environment is filtered by bilateral filtering of the optimized point cloud to achieve fast map construction.Providing critical a priori information for estimating and predicting the position and attitude of engineering vehicles.LiDAR-based local terrain environment reconstruction.Local terrain reconstruction of the environment in which the engineering vehicle is located is achieved using on-board LiDAR.To address the problems of data association failure due to unsteady movements of engineering vehicles driving on unstructured terrain and the difficulty of detecting and matching features in unstructured terrain environments.The priori motion information provided by the IMU is used to eliminate point cloud distortions due to non-linear unsteady motion,and then metrics are defined to assess local surface smoothness to characterize the edges and planes of points in the cloud.Further transforming the vehicle inter-frame motion estimation into a non-linear least squares problem based on the Levenberg-Marquardt(L-M)algorithm.The algorithm is based on the classical Lidar Odometry and Mapping(LOAM)algorithm and introduces a graph optimization theory based on Dynamic Bayesian Network(DBN)graph modeling in the G2O framework.The final result is a local reconstruction of the unstructured terrain environment,providing accurate information input for subsequent engineering vehicle position and attitude estimation based on matching local terrain with global map point cloud features.Position and attitude estimation of engineering vehicles based on a priori global terrain feature matching.The position and attitude estimation technique based on point cloud matching is investigated,and an engineering vehicle position and attitude estimation method based on a priori global terrain matching is proposed.Firstly,to address the difficulties of processing global point cloud data of large order of magnitude in unstructured terrain environments,the Soble operator for image processing is introduced based on the Gaussian kernel function estimation to obtain terrain point cloud edge information and combined with the Euclidean clustering algorithm for super voxel optimization to cluster non-ground points of terrain point clouds,and then the octree and cyclic voxels are used to optimize the storage,retrieval and management of point clouds to ensure that the obtained large to ensure the efficient management of the global point clouds obtained over a large scale unstructured terrain environment.Then,to address the difficulty of initial matching between global and local point clouds in a large scale,Principal Component Analysis(PCA)is introduced to reduce the dimensionality of the point cloud data to achieve the purpose of data simplification,and then the Normal Distributions Transform(NDT)algorithm and Iterative Closest Point(ICP)algorithm are integrated to achieve an efficient and accurate two-step matching of point clouds.The Gauss-Newton method is then used to solve the optimal transformation matrix to achieve accurate initial position and attitude estimation of engineering vehicles.Finally,to address the problem that the laser odometer position and attitude estimation algorithm is prone to cumulative errors when the ICP algorithm fails to match,residual compensation is introduced to correct the position and attitude estimation results and achieve accurate estimation of the position and attitude of the engineering vehicle.Position and attitude prediction of engineering vehicles in unstructured terrain environments.By analyzing the influence of unstructured terrain on the attitude of engineering vehicles,a method for predicting the position and attitude of engineering vehicles based on an a priori global terrain model is proposed.To address the problem that traditional methods have difficulty in accurately predicting the position and attitude parameters of engineering vehicles due to the neglect of the influence of terrain on the attitude of engineering vehicles.Firstly,the Newton-Euler method is introduced to construct an engineering vehicle dynamics model to resolve the coupling between ground and body attitude.The Genetic Algorithm(GA)was then introduced to improve the BP neural network by optimizing its weights and biases and then reassigning them to the neural network to accurately predict the future moment position of the engineering vehicle.The three-dimensional global terrain model of the engineering vehicle environment and the geometric model of the engineering vehicle are combined to obtain the vehicle tire contact point and the plane on which the body is located at the next moment,and finally to predict the attitude of the engineering vehicle at the future moment based on the constructed dynamics model.In summary,this dissertation constructs a global as well as a local 3D terrain model of the unstructured environment,and realizes the position and attitude estimation of engineering vehicles based on a priori terrain model matching,and further proposes a method for engineering vehicle position and attitude prediction.The research work of the dissertation provides a basis for the development of automatic driving and autonomous operation technologies for engineering vehicles,and provides technical support for the development of engineering vehicles in the direction of unmanned,intelligent and collaborative.