Research And Implementation Of Multi Robot Collaborative Mapping Algorithm Based On Point Line Fusion

Simultaneous localization and mapping(SLAM)enables mobile robots to autonomously locate and map in unknown environments.Among them,visual SLAM has become a popular research direction for SLAM due to its low cost and ability to simultaneously obtain color and spatial depth information.The visual SLAM system is already relatively mature in ideal environments,but in practical applications,relying solely on a single robot and point features for large-scale weakly textured structured indoor scenes can affect the efficiency and accuracy of mapping.In response to the above issues.This article proposes a multi robot mapping method that integrates point and line features for robot localization and mapping in indoor environments.This article is an improvement on the ORB-SLAM3 algorithm framework,and the main research content is as follows:(1)For some indoor scenes with weak textures and obvious features,this paper improves the visual odometer part of the ORB-SLAM3 algorithm and proposes a feature extraction algorithm that integrates point and line features.This algorithm adds line features to the original extracted environmental point features,that is,simultaneously extracting point and line features of image frames.The complementary information between line features and point features ensures the accuracy of visual odometer estimation in scenes with obvious weak texture structural features.(2)This paper proposes a map fusion method based on scene recognition to improve the efficiency of mapping for relatively large indoor scenes.This method relies on the visual word bag method to determine whether the scene area overlaps.Its essence is to determine the similarity between the current frame and the historical frame.However,if all the image frames collected by the robot are used for judgment,it will seriously affect the real-time performance of the entire system.Therefore,this paper proposes a keyframe extraction method based on photogrammetry to extract keyframes to improve the robustness of the algorithm.Finally,after a certain scene is determined to be overlapping,the P3P(Perspective 3 Point)algorithm and BA(Bundle Adjustment)algorithm are combined to solve the relative pose transformation relationship of multiple robots in the scene.Map fusion is performed based on the relative pose relationship and the implementation process framework of the map fusion algorithm is provided.(3)Build a software and hardware platform for mobile robots,transplant the improved multi robot joint mapping algorithm that integrates point and line features to this platform,and conduct experimental tests on public datasets.The experimental results show that the proposed algorithm has higher robustness and mapping accuracy compared to point based and line based algorithms.Experiments in indoor environments have also shown that the algorithm proposed in this paper performs well in weak texture and structured scenes,as well as in small-scale map fusion,achieving the expected goals.