Research On SLAM Technology Of Mobile Robots In Large-scale Indoor Scenes

With the continuous popularization and development of robot technology,the Simultaneous Location and Mapping(SLAM)technology of robots is also facing the challenge of solving more and more problems.For example,in the indoor environment of a larger scene,how to ensure the real-time and accuracy of the robot SLAM system processing massive data? In this paper,the SLAM technology of mobile robots in large-scale indoor scenes is researched,and the feasibility of each part of the algorithm is experimentally verified.Finally,the entire SLAM system algorithm is tested on the robot platform in the actual scene.map.The main research content of this article includes the following parts:First of all,this paper elaborated the mathematical model of the robot SLAM system in detail,and used this as the theoretical support to establish the observation model using the lidar as the sensor and the motion model with the robot odometer as the core.For the observation model,due to certain motion distortion during the movement,it needs to be corrected for distortion.The data obtained from the odometer motion model will have a certain amount of noise,so it needs to be calibrated.Through the preprocessing of data in the early stage,it lays the foundation for the research of subsequent algorithms.Secondly,this article has conducted detailed and in-depth research on the various modules of laser SLAM in the large scene.In larger scenarios,the accuracy and real-time performance of the front-end data processing will directly affect the performance of the entire system.Therefore,on the basis of establishing a spatial likelihood domain model,this paper uses a CSM-based matching algorithm to position the robot.The initial pose is estimated,and the branch-and-bound algorithm is applied to ensure the accuracy of the matching algorithm.Afterwards,a method for adaptively determining the search range is proposed,which improves the setting of the search range and greatly improves it.The search rate of the algorithm.Again,for the increasing cumulative error,this paper uses a loop closed detection algorithm,adding additional constraints on robot pose estimation,and again uses the branch and bound method for accelerated matching during the closed loop detection phase.After that,a graph model is constructed for a series of poses of the robot.Based on the g2 o library,a nonlinear optimization method is used to perform graph optimization experiments on the pose image.The experiments used Intel and Killian data sets to verify the feasibility of the optimization algorithm.Finally,in order to verify the performance of the overall SLAM algorithm,in this chapter,through the actual robot experiment platform turtlebot2,equipped with lidar,the experiment was carried out in the experimental teaching building,a grid map of the experimental floor was established,and this algorithm was compared with the mainstream robot laser SLAM algorithm Gmapping was compared,and the entire experimental results show that the algorithm in this paper can be applied to a large range of indoor scenes in terms of system real-time performance and accuracy.