Multi-sensor Fusion Positioning Based on Prior Map Constraints
The core foundation for the autonomous execution of tasks by mobile robots lies in their localization tech-nology.Through precise positioning,mobile robots can effectively carry out tasks such as path planning and control,and target detection.Aiming at the significant accumulative error,low accuracy and poor real-time performance of ground mo-bile robot positioning algorithms in large-scale complex scenes,a positioning algorithm based on a priori point cloud map constraints integrating multi-line LiDAR,inertial measurement unit (IMU) and GNSS is proposed.Initially,the robot pose is initialized using the GNSS information and key frames saved in the prior map with the currently acquired GNSS measure-ment values and laser frames.Subsequently,the IMU is pre-integrated to estimate the robot pose,and a local map is con-structed using the adjacent key frames of the estimated pose.Finally,this local map is registered with the current laser frame,thus achieving precise positioning of mobile robots.High volumes of experimental results from both public datasets and real scenes demonstrate that,compared with existing algorithms,the proposed algorithm improves accuracy by about 20％ and localization speed by about 1％.The multi-sensor fusion positioning algorithm based on prior maps provides high-er accuracy and reliability in large-scale complex scenes,and can be used as a practicable positioning solution for mobile robots in a wide range of large-scale settings,from rural and forest scenes to urban areas.
NETL
NSTL
万方数据
