Research On SLAM Algorithm For Multi-source Perception Fusion Of Mobile Robot

With the development of computer vision technology and artificial intelligence technology,Visual SLAM algorithm is currently widely used in autonomous positioning and mapping on mobile robots,but how to achieve and deploy the corresponding Visual SLAM algorithm for actual application scenarios is the top priority of SLAM research.Based on the Visual SLAM algorithm and integrating multi-source information input,this paper introduces a kind of Visual SLAM algorithm of multi-source perception fusion for ground mobile robot.The main work of this paper is as follows:1.Multi-visual feature tracking algorithm based on point and line feature.Line features are added to the existing visual slam front-end algorithm based on point features to improve the robustness and accuracy of visual SLAM Algorithm in man-made structural environment;based on the analysis of open source line feature extraction and matching algorithm,a improved EDLine algorithm is improved to improve the integrity of line feature extraction;a line matching algorithm based on optical flow tracking is designed to improve the matching speed of line features.2.Multi-source information fusion odometry algorithm for mobile robot.According to the motion model of IMU(inertial measurement unit)and wheel odometer,a joint pre-integration algorithm of IMU wheel odometer is proposed to reduce calculating consumption of fusion odometry algorithm;a fast initialization algorithm of visual SLAM Based on IMU and wheel odometer data is designed;a back-end optimization algorithm based on the visual point line characteristics of mobile robot platform and IMU wheel odometer is proposed to optimize the robot pose estimation.3.Closed loop correction algorithm for multi-source information fusion.Based on the existing research results of image retrieval based on deep learning,a closed-loop detection algorithm using IMU wheel odometer data is proposed,and a set of pose map optimization algorithm combined with IMU wheel odometer data is designed.4.Integrating the algorithms of the above three modules,a set of integrated slam system is designed and implemented,which is verified on the ground mobile robot and the public dataset.