Application Research Of Indoor Mobile Robot System Based On SLAM

With the sweeping robot based on SLAM(Simultaneous Location and Mapping)coming into everyone's life,and the intelligent robot technology is gradually understood by everyone,which indicates that the robot related technology is more and more mature.The research of robots based on deep learning,target detection and semantic maps has promoted the development of intelligent robots and has important theoretical and practical significance.This paper has carried out the following work in the application research of SLAM-based mobile robot system.Firstly,a 2D laser SLAM-based robot was designed and built.Combined with ROS(Robot Operating System),the robot's function development mode and communication method were standardized encapsulated.Secondly,the Gmapping algorithm based on Rao-Blackwelized particle filter,the charting principle of Cartographer based on graph optimization,and the mapping effect of these two algorithms in the scene with corridor environment are studied.The advantages of Cartographer's loop detection are compared and analyzed.The feasibility of multi-machine collaborative mapping is analyzed through Gazebo simulation environment experiment.Thirdly,several global path planning algorithms in robot navigation are discussed,and the efficiency of A* algorithm is verified by comparison.This thesis studies the real-time local path planning algorithm DWA(Dynamic Window Approach)and its path planning principle,which embodies better real-time obstacle avoidance ability.In the end,the navigation function is discussed,and the performance of AMCL(adaptive Monte Carlo localization)localization algorithm is validated for the kidnapping problem.Fourthly,a localized AI scheme based on the Raspberry Pi equipped with Realsense D435 RGBD camera,neural computing stick and target detection algorithm MobileNet SSD model trained by Caffe deep learning framework was designed and implemented.This scheme enables SLAM robots to add semantic information to two-dimensional maps in real time.Aiming at the conflict between traditional high-precision semantic labeling strategies and semantic navigation,a semantic labeling strategy based on action area is proposed,and an actual and effective semantic navigation experiment is carried out with an application case of autonomous control of air conditioning by robots.Finally,the SLAM robot is applied to the housekeeping service scenario,and the two application scenarios of the data remote communication system based on the wide area network and the wireless ROS robot monitoring system based on the customized Android system are considered respectively.Among them,the remote communication system enables the housekeeping family service robot to report the household environment data regularly,and can monitor whether someone stranger enters the home without anyone at home,and provide the stranger's photo to the household owners.In the case of abnormally broken routers,in order to ensure communication stability,consider the mechanism for uploading through GPRS and NBIOT.The Android-based wireless robot monitoring system uses Android mobile devices to communicate with ROS robots through Rosbridge.