Indoor Map Construction And Navigation Of Mobile Robot Based On Multi-sensor Fusion

With the extensive use of sweeping robot,indoor mobile robot has gradually become the research focus in the field of artificial intelligence.They have been involved in industrial automation and daily life.The guiding robots in shopping malls and hotels and the sorting robots in libraries are all indoor mobile robots.They can greatly reduce people's workload and improve people's lives.Based on the background of AMR(Autonomous Mobile Robot),this paper mainly studies the problems of indoor environment map construction and autonomous navigation with building a robot system combining software and hardware.This paper proposes a robust indoor map construction scheme from the perspective of sensor fusion,which can expand the map accurateness and navigation effectiveness.The main work of this paper is as follows.The paper researches and builds a robot system that can run the mapping and navigation algorithm.According to the application scenario of indoor mobile robots,the Raspberry Pie 3B+microcomputer based on Linux system is used.The dual wheel differential drive is used to realize the small and light appearance.Crawler type wheels can prevent collision.The robot obtains external information with camera,lidar,inertial measurement unit and other sensors.At the same time,the mapping and navigation software system is built under the framework of ROS(Robot Operating System)to ensure the reliable transmission and accurate processing of data.A loop detection method in multi-ring complex terrain is proposed in this paper.The paper discusses the method of obtaining environmental point cloud map by lidar and camera,including grid map generation,scan matching,key frame extraction,pose estimation and so on.In order to improve the efficiency and accuracy of loop detection,the BoW(Bag of Words)model is used to detect the loop in multi-ring complex terrain,and an image dictionary is trained to fit the indoor environment.The monocular camera is accustomed to obtain the environment image for ORB feature detection and extraction,and the prior similarity method is used to improve the robustness of the loop module when calculating the frame similarity.This paper presents a mapping and navigation scheme based on the fusion of lidar and camera data.Simple and cheap monocular camera is selected to provide visual information,together with lidar to build the environment map.This can be applied to a wider scene.In order to improve the problem of loop misjudgment,a 3D point cloud mapping method is used to generate 2D grid to correct the grid map.At the same time,autonomous navigation is carried out on the basis of accurate map,and A*algorithm based on dynamic measurement heuristic is adopted to improve the efficiency of finding target points.The navigation efficiency of robot in actual situation is improved through inflection point optimization.