Monocular Simultaneous Localization And Mapping System Based On Binary Visual Features

Recently,mobile robots have been widely used in the aerospace,military,rescue and domestic service areas,and it has become one of the hottest topics in the field of Robotics.In particular,the Simultaneous Localization and Mapping(SLAM)is considered as one of the key components to realize the autonomous operation of mobile robots.However,there are still many challenging problems for mobile robots to realize higher intelligence,and a number of key technologies in SLAM remain to be solved.This thesis focuses on the monocular SLAM based on binary features and multi?robot cooperative SLAM.This thesis includes the following three points:First,this thesis summarizes the various SLAM methods currently used in mobile robots,and analyzes the characteristics of the sensors they use.Consequently,we choose the monocular vision as the perceptual approach in the proposed SLAM system.At the same time,according to the characteristics of outdoor application environments of mobile robots,this thesis also analyzes the description methods of the environments and chooses the local binary visual features for the proposed SLAM system,considering their good performance in fast calculation and small storage occupation.Secondly,according to the existing problems of the monocular SLAM running in outdoor large?scale environments,this thesis proposes several corresponding improvement schemes.1)Concerning the tracking failure problem,a relocalization method based on active loop closure is proposed to improve the robustness of the monocular SLAM when the robot explores unknown environments.2)Concerning the problem of limited computation and storage capacity of mobile robots,a distributed multi?robot cooperative SLAM system based on monocular vision is proposed.This system divides the large?scale outdoor environment into several small areas and uses multiple robots to explore each area at the same time.3)Concerning the problem that the monocular SLAM has no real scale of the environment and the map is less impressive to fulfill the robotic urban search and rescue tasks,A combined SLAM system based on monocular vision and 2D LiDAR is proposed.it can solve the problem mentioned above and provide a more understandable map together with a automatic victim recognition module to the operators.The system has been successfully tested and applied in RoboCup Rescue Robot League competitions.Finally,based on the NuBot mobile robot developed by our group and the famous KITTI dataset,a series of indoor and outdoor experiments are carried out,and the performance of the proposed SLAM system is analyzed both qualitatively and quantitatively.The experimental results show that the proposed monocular SLAM system based on the binary features effectively improves the robustness and accuracy of the mobile robots' localization and mapping,and an initial implementation of the multi?robot cooperative SLAM system has been realized successfully,which lays a good foundation for achieving a higher level of the autonomy for mobile robots.