Research On Robust Map Construction And Small Obstacle Measurement Algorithm Based On LiDAR

Li DAR has the characteristics of high measurement accuracy and strong anti-interference ability,and is an important sensor in the field of robot technology.Constructing an environmental map based on Li DAR and performing accurate measurement of obstacles are the basic tasks to achieve autonomous driving(walking)of robots.However,the existing algorithms still have limitations: First,the current laser SLAM algorithm ignores the relationship between laser point cloud information,Resulting in incomplete drawings and time-consuming processes in common scenarios such as corridors and intersections.The second is the use of visual information to measure the accuracy of small obstacles is generally not high,and the laser point cloud information density is low,it is difficult to meet the measurement requirements.Therefore,how to use Li DAR to efficiently complete map construction and how to complete accurate measurement of small obstacles is a very research topic.This can not only greatly expand the application scenarios of Li DAR,but also provide advanced path planning and dynamic obstacle avoidance for robots.Provide helpful help in decision-making.Based on the above background,this thesis takes laser point cloud information processing as the main content of the research,and conducts research from the construction of robust maps based on Li DAR and the accurate measurement of small obstacles.The specific research contents are as follows:(1)A map construction algorithm in the non-echo region of Li DAR is proposed,which improves the efficiency of map construction and makes up for the deficiency of laser SLAM completely relying on echo data for map construction.Perform clustering and information-type endpoint feature extraction on the laser point cloud information of the echo area to determine the echo-free area,and then update the grid occupancy probability of the area to construct a complete map.The actual test was carried out with the help of the Robot Operating System(ROS).The experiment shows that the algorithm can accurately complete the mapping task of the echo-free area,and the maximum time can be saved when constructing the local corridor map(10m × 2.5m)In 183.6s,the map construction area was increased by 56.7%,and the map construction error was less than 1.2%.(2)An obstacle measurement method based on 3D laser point cloud information is proposed.Comprehensive use of the distance and angle information of the 3D laser point cloud information to construct a reasonable mathematical model to complete the measurement of the obstacle's maximum width,distance and azimuth angle.Using a robot as a carrier,using 16-line Velodyne Li DAR for outdoor experiments,when the distance from pedestrians ranges from 4.51 m to 7.50 m,the average errors of distance measurement and maximum width are 0.32% and 2.11% respectively,and the error of azimuth angle measurement is 0.20.When the distance range is 14.56m?17.54 m,the average relative errors of distance measurement and maximum width are 0.35% and 2.28% respectively,and the error of azimuth angle measurement is 0.23.(3)A small obstacle measurement algorithm that combines Li DAR and visual information is proposed.The measurement accuracy of using a single sensor is low,the efficiency of the existing fusion algorithm is low,and it is difficult to take advantage of each sensor when measuring small obstacles.Therefore,a decision-level fusion algorithm of "Li DAR verification vision" is proposed.First use visual information to detect obstacles based on deep learning.During the robot movement,when the amount of laser data in the detection area meets the requirements,project the visually detected area onto the Li DAR coordinate system to obtain laser point cloud information in the current area.Furthermore,small obstacles are measured based on the obstacle measurement algorithm.Using 16-line Velodyne Li DAR and industrial IDS camera for method verification,the results show that even if the method is used in the process of robot movement,the maximum width error of measuring small obstacles can be guaranteed to be less than 2.4%,and the ranging error is less than 0.15%.