Research On Mobile Robot SLAM Based On Multi-sensor Technology

In recent years,with the increasing popularity of mobile robots,artificial intelligence,unmanned driving,and computer vision,more and more technical difficulties have been overcome by scientific researchers.SLAM,as the difficulty in navigation research of mobile robots and the cornerstone of driverless technology,has attracted more and more scholars at home and abroad.The mobile robot SLAM is a problem that must be solved to achieve intelligent robots.Comprehensive consideration of the current SLAM method,relying on a single 2D laser sensor,has limited perception of environmental information;relying on binocular vision sensors,the real-time performance is relatively poor.Depending on the depth camera,it is easy to be affected by light intensity.Therefore,this paper focuses on the method of location mapping based on the fusion of multi-sensor technology to increase the awareness of the environment,improve the adaptability of the mobile robot working environment,and then increase the robustness of the navigation system.This is also an important direction for the development of mobile robot SLAM.The main contents of this paper are as follows:Firstly,the mobile robot model,laser sensor,depth camera,and odometer model are established.The models and basic parameters of the laser radar and depth camera described in this paper are introduced,and the depth camera is calibrated.Secondly,2D laser radar SLAM method based on the particle swarm algorithm,Gmapping algorithm is analyzed.Then the resampling algorithm in particle filter is studied emphatically.The problem of particle diversity and particle degradation in resampling is analyzed and an improved hierarchical genetic variation is proposed.Sampling particle filter algorithm,in the MATLAB simulation environment,compares the improved algorithm with the current mainstream resampling algorithm to illustrate the effectiveness of the algorithm described in this paper.Again analyzed the depth camera SLAM feature point detection method of the current mainstream in choosing a higher real-time ORB detection algorithm,and then studies the current mainstream RANSAC algorithm used for outlier elimination.To solve the problem of high algorithm complexity caused by the initial random selection of sample points for RANSAC algorithm and the algorithm complexity caused by the removal of all samples is proposed,a RANSAMC algorithm was proposed,and the effectiveness of the algorithm described in this paper was verified through simulation and comparison experiments.In the end,the fuzzy logic decision-level fusion 2D laser radar and depth vision are used to realize the construction and fusion of the grid map based on three-dimensional environmental information,and the physical experiment verification of the mobile robot BLACK-2 platform carried in the mechanical electronic laboratory.