Study And Application Of Road Environment Perception System For Unmanned Mining Truck

Due to the harsh environment of the mine,the application of the unmanned driving system in this scenario needs to be implemented urgently.The mine unmanned driving system not only ensures the safety of employees,but also improves work efficiency.Among them,the perception of the mine’s road environment is a prerequisite for the successful operation of the mine unmanned driving system.Aiming at the demand of mine unmanned driving system,this thesis uses mining trucks equipped with multiple sensors to build a road environment perception system applied in mine scenarios.The main research is as follows:(1)Investigate the processing algorithm of sensor data.Firstly,the mine environment is scanned by using the laser radar mounted on the mining truck;secondly,the distortion correction of the point cloud is carried out by using the IMU data corresponding to the time stamp,and then the point cloud is denoised by various filtering methods;finally,the RTK pose information processing is studied,the RTK pose information is fused with the IMU and laser odometer information to realize the vehicle positioning when the pose information is lost due to the RTK signal being blocked,and to provide stable pose information for the subsequent perception algorithm.(2)The boundary line extraction algorithm of mine roads is studied.Firstly,ground segmentation is performed on the denoised and de-distorted lidar point cloud to obtain the ROI area;secondly,the road boundary coordinate points are selected in the ROI area by using the velocity and angular velocity information of the mining truck combined with the structural characteristics of the mine road;after the coordinate points of the road boundary line,optimize the coordinate points such as smoothing and interpolation;finally,convert the coordinate points of the boundary line from the spatial rectangular coordinates to the high latitude and longitude coordinates to improve the versatility of the road boundary line information.(3)A filtering algorithm for removing redundant point clouds using the extracted mine road boundary line and an algorithm for detecting whether there is obstacle encroachment in the safe driving area of mine roads are studied.Firstly,use the Gauss-Newton method to fit the obtained left and right boundary lines with a piecewise quadratic function,and then use the corresponding function relationship to complete the filtering of the lidar redundant point cloud and extract the road point cloud;secondly,use the improved RANSAC algorithm and the DBSCAN clustering algorithm is used to extract the point cloud of road obstacles;the safe driving area and the obstacle point cloud are rasterized to generate a two-dimensional grid map,and the state of the grid is used to judge whether there is an obstacle encroachment in the safe driving area;finally,output the encroachment information and visualize it in Rviz.(4)Deploy the algorithm on the mining card platform and test it in the actual application scenario of the mine.When the RTK pose information is temporarily lost,test whether the fusion algorithm can provide pose information,and analyze and evaluate the accuracy of the pose information.After testing,the average error between the fusion pose and the true value is only 0.03m;The boundary line extraction algorithm is tested in three mine scenarios,and the experiments show that the algorithm has achieved good results in these three road conditions;in the real road scene,the obstacle encroachment detection algorithm is tested,and the results show that the algorithm can be compared Accurately reflect the encroachment of obstacles in the safe driving area.The road environment perception system constructed in this project covers multiple aspects such as sensor data processing in mine scenarios,road boundary extraction,and obstacle encroachment detection in road safety driving areas,laying a certain foundation for the application of unmanned driving systems in mines.