Research On SLAM Technology For Mobile Robot In Unknown Environment

Autonomous mobile robot is an important and hot spot in robotics research.In order to achieve the autonomous navigation accurately,robot has to know its localization in advance,especially when it doesn't have map beforehand.In the unknown environment,in order to complete the work of further navigation,the first work for the robot is to build the environment map through its own sensors.The mapping and localization for the robot in unknown environment is proceeded at the same time.Therefore,Simultaneous Localization and Mapping(SLAM)has become one of the fundamental and hottest problem.Both Laser scanner and Kinect has been widely used in robotic application for 2D Simultaneous Localization and Mapping.The map whether to be used for navigation due to the accuracy of a map but the map accuracy is limited by the field of view of the sensors.This paper firstly introduce the technical basis of SLAM and the main achieve methods of SLAM.Secondly,introduce the robot platform,the sensors used by the robot and the Robot Operating System(ROS),the ROS is an open source robot operating system.It comes with a visual and a real time dynamic parameters application interface.Users can check the system status and test the packages running by using the commands in ROS.Thirdly,analyze the crucial problems of the SLAM technique,such as map representation,description and process of uncertain information,localization,data association and accumulative error.Rao-Blackwellized Particle Filter is used in the localization algorithm,the principle and steps of Rao-Blackwellized Particle Filter are analyzed in detail.Finally,principally apply the Laser scanner and Kinect,four methods of sensor patterns for SLAM are used: a single Kinect,two Kinects,a Laser scanner,a Kinect combine with a Laser scanner.Three SLAM algorithms include Gmapping,Hector SLAM and CRSM SLAM are tested using the four methods to build accurate 2Dmaps for comparison.All the methods have been evaluated and compared in real world experiments with slight,less and complex features,then the performance of the three SLAM algorithms were compared particularly in the map accuracy.