Research On Traversability Of Mobile Robot In Indoor Environment Based On SLAM Technology

Location and navigation of indoor robots is one of the key technologies in the field of robotics research.Traditional methods based on ultrasound,infrared switch,encoder and so on can only provide limited data,and can not effectively perceive the environment.In recent years,the development of SLAM technology has provided a new solution for robot localization and mapping of its environment.However,the complex terrain and various obstacles in the actual scene have a great impact on the measurement of visual signals.At the same time,visual SLAM technology relies on a large number of image data and complex computation,which makes the robot control lack real-time performance and high error rate.In serious cases,it will lead to the loss of map information.Therefore,there are still many difficulties in real-time control and navigation of robots using SLAM system constructed by vision.In this paper,a fast method of building compact map based on visual SLAM system is designed for robot navigation in complex indoor environment.Based on the traditional RGBDSLAM algorithm,this paper simplifies the map to a 2D raster map,and designs an independent path search algorithm to control the navigation of the robot based on the known map.This method effectively solves the navigation and obstacle avoidance problem of the complex indoor space robot.Finally,a complete experimental system is constructed based on the existing robot platform in the laboratory.The method in this paper first establishes a front-end visual odometer module,which is used to estimate the unknown and global motion modules of the robot with the back-end optimization module.At the same time,a global point cloud map is generated to record the indoor environment that the robot has perceived.Then,combined with the current pose data of the robot,the global point cloud image is cut into a smaller area in front of the direction of the robot motion.At this time,the volume of the data is reduced from more than 10 million threedimensional points to more than 1 million three-dimensional points.Finally,according to the actual volume of the robot,the cut map is sampled as a 2D raster map.Based on these features,a Gauss model is constructed to determine the final traversability parameters,and a traversability map is obtained.Compared with the original data,the final map data is compressed by several orders of magnitude,and its size is only more than 1500 twodimensional points.This paper creatively puts forward the concept of accessibility parameter,which can be obtained by pattern recognition.The raster map of a small area is fitted into a plane,and the characteristics of the plane are extracted to judge the accessibility parameters.This method greatly simplifies the size and complexity of the map.In order to verify the reliability and realtime performance of the accessibility parameters,this paper tests the navigation effects of two path planning algorithms in unknown environments.Finally,this paper constructs a mobile robot platform,carries out large closed map experiments on indoor flat ground to verify the validity of SLAM technology.According to different terrain,combined with the strong trafficability of tracked robot,experiments on trajectory planning and obstacle avoidance are carried out,which proves the feasibility and validity of SLAM method and trajectory planning using trajectory maps in complex indoor environment.