Research On Real-Time Positioning And Mapping Technology Of Intelligent Vehicles Based On Multi-Sensor Fusion

SLAM(simultaneous localization and mapping)technology refers to the intelligent vehicle carrying various sensors to estimate and map its position while moving under the condition of uncertain position.Because the single-line lidar lacks altitude information and can’t scan obstacles below the sensor,the mapping is inaccurate.However,visual pairs are easily influenced by illumination,and it is challenging to identify characteristics when there is a lack of detailed environmental data,which can result in feature tracking difficulties in real-world scenarios.In this paper,inertial measurement unit,depth camera and lidar data are fused to build an experimental platform of smart car based on ROS,which realizes real-time positioning and mapping of smart car in unknown environment.The main research contents are as follows:Firstly,the motion model of intelligent vehicle is established,the relevant principles of each sensor are analyzed,the camera distortion is corrected and the internal and external parameters are calibrated.Secondly,the real-time positioning and mapping technology of lidar and vision camera is studied and analyzed,and the mathematical model of SLAM is established.Two common algorithms of real-time positioning and mapping technology of lidar are simulated and analyzed,and the typical algorithm of visual SLAM is analyzed and verified,and the transformation from point cloud map to grid map is completed.Thirdly,the research part of laser radar and camera fusion positioning and mapping is analyzed,the commonly used algorithms are simply enumerated,the advantages of multi-sensor fusion are introduced,the laser radar and depth camera are calibrated jointly,and the rotation matrix and translation matrix are obtained.Combining the advantages of lidar and depth camera,this paper puts forward a method of constructing a map by multi-sensor fusion,adopts extended Kalman filter to fuse its attitude,and adopts Bayesian method combined with map fusion rules to fuse local raster maps,which is verified by simulation on MIT data sets.Finally,taking Autolabor Pro1 intelligent vehicle as the experimental platform,the experiment is demonstrated by installing two-dimensional lidar,inertial navigation,depth camera and other sensors before and after the mobile chassis.The results show that the positioning and mapping method based on multi-sensor fusion is more complete and accurate,and the navigation robustness of smart car in unknown closed environment is improved.