Autonomous Navigation Of Mobile Robot Based On Information Fusion In Indoor Environment

With the continuous development of science and technology,the intelligentization of robotics technology has become a research key point,and slowly approaching people's daily lives.Simultaneous Localization and Mapping and real-time path planning are key technology for intelligentizing mobile robots.The accuracy of the map directly affects the localization and autonomous navigation of the mobile robot,so how to improve the accuracy and integrity of the map is crucial to the robustness and real-time performance of the autonomous navigation of the mobile robot.In this paper,the object of the study is the family service mobile robot,which had a certain requirement for the cost of hardware.In the experiment stage,the low-cost lidar was used to fuse visual information.On the basis of reducing the development cost,the accuracy and integrity of the map were effectively improved.Therefore,based on the Nation Natural Science Foundation Item “Online Verification of Vehicle Autonomous Decision-making Safety Based on Formalization and Ad-Hoc Method”,the main contents and work in this thesis are as follows.1.Based on the ROS system,I built mobile robot platform with Rplidar A2 and Tuetlebot2 and Gmapping to realized localization and map building of mobile robot in static environment.But RpidarA2 as a single line laser radar,the scanning area is a two-dimensional plane.The obstacle that is out of the two-dimensional plane is not scanned,which will affect the obstacle avoidance of the mobile robot.Therefore,the three-dimensional information of the obstacle was added to the two-dimensional grid map of lidar scanning to enrich the information of the map and improve the accuracy of the map.2.Using the Microsoft's RGB-D depth camera Kinect to do visual SLAM experiments,and to analyze the limitations of visual SLAM for map building.In visual SLAM,the point cloud where the camera acquires the environment is very large.If it is used directly,the amount of computation is very large,the memory and CPU requirement of device are high.From the extraction point cloud to the use of feature point method to set up VO(Visual Odometry),set the threshold value,filtered the point cloud beyond a certain range,and improved the calculation speed.3.In this paper,the indirect method Bayesian estimation was used to fuse the data acquired by single-line lidar and RGB-D camera,and a light projection model is established.The light beam was simplified and abstracted into a certain distance light torecord grid cells,according to occupy the probability of grid to gradually establish a grid map.That improved accuracy of the map and ensured integrity of the map.At the same time,the accuracy and real-time performance of obstacle avoidance and navigation of mobile robots were indirectly improved.4.To build experimental platform of mobile robot has carried on the study and research,mainly includes the hardware the mobile platform Turtebot2,sensor single-line lidar Rplidar A2 and RGB-D camera Kinect,the software ROS(Robot Operating System).Using the experimental platform to do the experiment of map building and navigation in the actual scene,and proved the feasibility of this system.