Research And Implementation Of SLAM Technology For Indoor Surveying And Mapping Robot

In recent years,with the wide application of artificial intelligence in various disciplines,robotics has realized a leap from theory to practical application.The SLAM algorithm is an important research direction in the field of robotics.It is the basis for the robot to autonomously navigate and complete complex intelligence tasks in an unknown environment,and is also an important basis for the robot's perception ability and intelligence level.Surveying and mapping robots are the new development direction of the surveying and mapping industry and have great significance for the realization of surveying and mapping's automation,informationization and intelligence.Due to the influence of signal occlusion and other factors,traditional GPS technology can't perform indoor localization and measuring.In view of this problem,this paper independently designs the mapping robot system and applies SLAM technology to realizes automatic drawing of two-dimensional indoor maps and navigation and positioning functions.The main research content is as follows:(1)Acquire indoor environmental information by using sensors such as laser lidar,wheel odometer,and IMU on the mapping robot platform,and use the Hector SLAM,Gmapping,and Cartographer algorithms to perform real-time indoor mapping experiments for the mapping robot.The results show that the three algorithms can obtain indoor two-dimensional maps in real time.Compared with the traditional measurement,it can save human resources and time costs.(2)Through the comparison and analysis of the three SLAM algorithms,the results show that the Cartographer algorithm's mapping accuracy is obviously better than the Hector SLAM and Gmapping algorithm,and the mapping accuracy reaches 5cm,which can meet the basic indoor service requirements;meanwhile,because of its closed loop capability makes the Cartographer algorithm very robust.(3)This paper innovatively uses the deep learning method to solve the robot localization problem.Through the training and learning of robot localization process,and used FFNN and CNN neural network models to complete the position of robot.The results show that the neural network is effective and feasible for robot positioning.It also shows that deep learning has great potential in robot localization.