Research On Vehicle Multi-Feature Fusion Lidar Localization Method In Structural Driving Environment

The current rapid development of autonomous driving technology,which some head enterprises have already applied in real.Real-time and accurate localization is the key to the operation of self-driving vehicles.Only finding out the vehicle’s location,such as planning,controlling and many other tasks can be completed in order.The traditional positioning method by GPS is prone to lose signal in a dense environment with tall buildings,which caused the localization accuracy can not be guaranteed.In this thesis,I propose a localization algorithm that relies on the fusion of scene geometric structure information and lidar reflection intensity information to improve the robustness and accuracy of localization.The main contents of this paper are as follows.Firstly,as for the problem that previous lidar localization algorithms seldom utilize the ground point cloud data,an algorithm is proposed to fit the ground plane and identify road curbs.In detailed,it uses a method of calculating the relative slope and RANSAC(random sampling consistency)to extract the ground effectively in real time.A sliding window algorithm is proposed to detect road curbs to improve the safety of autonomous driving.Secondly,as for the problem that lidar localization relies too much on scene geometry information while the information channel of reflection intensity is not utilized.A map generation algorithm is proposed to generate a reflection intensity map of the ground,according to the feature that reflection intensity of all kinds of materials are different.During the locating process,a SLAM algorithm is used to calculate poses in real time and an improved image matching method is proposed to match and calculate the local intensity map with the global intensity map to correct the initial position,which effectively improves the robustness and accuracy of the lidar positioning algorithm.Finally,the localization procedure is tested and evaluated by a mobile platform to analyze the localization errors with the campus data.