Design And Research Of The Automatic Driving Sensing System Of Trams Based On Multi Source Heterogeneous Information Fusion

Research on the application of autonomous driving technology to rail transit has gradually increased with the rapid development of urban rail transit.On the one hand,due to its single coverage and poor robustness,a single sensor is difficult to meet the complex sensing environment faced by rail transit automatic driving;on the other hand,due to the unique track boundary characteristics of rail transit,it can accurately determine obstacle targets Whether or not to invade the restricted track is very important to the safe driving of autonomous trams.However,the traditional method of using cameras to identify the track relies heavily on the environment,light and weather,and the identification distance cannot meet the requirements.This paper is mainly based on the application scenarios of modern trams,builds a testing platform for the perception system,and carries out research on the detection method of obstacles in the track based on multi-radar and GPS/INS positioning,which can provide in-track for the decision-making control system of the automatic driving of trams.The status information of the obstacle target.The tram automatic driving system designed in this paper uses perception information and computer technology to enable the decision-making control system to obtain a reasonable operation plan,so that the tram improves the comfort of passengers on the premise of improving the safety of operation.The main research contents are as follows:(1)Aiming at the installation error of the multi-radar system on the platform,a method of laser radar coordinate correction for trams is provided.In the point cloud ground point filtering,a multi-iterative ground fitting method is proposed to solve the problem of the inconsistency of the ground accuracy of the angle method.On this basis,the non-ground point cloud is rasterized and down-sampled.(2)Using the scan line characteristics of the real-time point cloud of lidar,and aiming at the high computational cost of traditional density clustering methods,a clustering algorithm based on scan lines and horizontal angles is proposed.The clustering result is modeled by the minimum rectangle envelope method,and the obstacle information is associated and fused for the obstacle target list generated by each sensor.(3)According to the high-precision latitude and longitude information and heading information provided by the GPS/INS inertial navigation equipment,the route map is recorded,and the track area in front of the vehicle is obtained in real time under the running state of the vehicle,and mapped to the rasterized real-time map to form the region of interest,which is in the ROI Screen the invading obstacle targets and send them to the decision-making control system to provide a basis for the automatic driving system of the tram to issue a warning or take a brake.(4)A modern tram test platform based on lidar,millimeter wave radar and GPS/INS inertial navigation equipment has been built to analyze and process the data of various visual sensors and combine GPS/INS inertial navigation positioning information to provide a sense of forward track for the decision-making control system Obstacle information in the area of interest.After the on-site loading test in the test section and the operation section,the experimental data was collected and the experimental results were analyzed.The experimental results prove that the running results of the related algorithm of the tram perception system designed in this paper can achieve the output of the intrusion limit obstacles in the track,and the effect is in line with expectations.