Application Of SLAM Algorithm Based On Stereo Vision In Omnidirectional Mobile Robot

Intelligent mobile robot in an unknown environment,do not rely on external positioning signal,through its own sensor to achieve pose estimation and incremental build map,and then get their own positioning,this process is called SLAM(Simultaneous Localization and Mapping).The SLAM problem is one of the research hotspots in the robot field at present,and has important theoretical significance and application value for realizing the robot intelligent.In this paper,the SLAM algorithm based on stereo vision is studied firstly.The RGB-D camera is used as the SLAM sensor to construct the 3D map and locate the space.The SLAM algorithm based on stereo vision consists of four parts: VO front end,back end optimization,loopback detection and mapping.The VO front end uses stereo vision to estimate the robot's motion state.It uses ORB features to extract and match the features.The Pn P problem is used to estimate the motion transformation.The ICP algorithm is used to efficiently segment the 3D point cloud.The backend uses graph optimization algorithm to establish the same global map and track of VO front-end data.Use the loopback detection algorithm to further improve the accuracy of the map.The use of Octo-map to build the map,the storage structure of its small footprint,and can handle moving objects and for navigation.Then an omni-directional mobile robot was designed and built.It has an omnidirectional mobile chassis based on Mecanum wheel with a lidar,depth camera and a four-degree-of-freedom arm.Its computing core is Nvidia Jetson TX2 high-performance embedded computing unit,powerful,compact,energy efficient.Finally,the stereoscopic SLAM algorithm described in this paper is validated by the constructed omnidirectional mobile robot,and is operated in a real environment to reconstruct and locate the three-dimensional map.The real-time of the algorithm and the validity of the system architecture are proved,and the results are analyzed.