Research On Autonomous Nacigation Method For Robots Based On Visual-inertial Fusion

Detection task in the variable indoor environment is an important application of mobile robot in the field of rescue and relief.It can complete point-to-point detection task in the environment,transmit the information inside the disaster environment to rescuers,and assist in the rescue task.Therefore,it is of great significance to study mobile robots with the abilities of environment perception,decision judgment and behavior execution,and SLAM(Simultaneous Localization and Mapping)and navigation technology are the technical core and key for mobile robots to accomplish this task.In this paper,a SLAM and navigation system equipped with binocular camera and IMU is designed and implemented according to the requirement analysis of the exploration task.Firstly,the requirements of point-to-point detection task in the variable indoor environment were analyzed,and the modular design of SLAM and navigation system was carried out to ensure the scalability of the system.The two subsystems can run independently and communicate through the mechanism of the message bus.SLAM subsystem realizes positioning and mapping of mobile robot in unknown indoor environment.Firstly,the initial motion estimation of the camera is obtained by comparing the adjacent frame image information collected by the camera with the optical flow tracking method,and then the IMU deviation term is incorporated into the optimization term for joint optimization,so as to obtain a more accurate motion estimation result.At the same time,an incremental environment map is created based on the image depth information and motion estimation information.The navigation subsystem realizes the safe advance of the mobile robot to the target detection point to carry out the task of environment detection.Dijkstra orA~*algorithm is used to plan the global optimal path from the current position to the target position.In the process of moving along the track,continue to perceive the surrounding environment,and use the improved DWA local path planning algorithm to avoid dynamic obstacles and reach the target safely.Finally,the SLAM and navigation system designed in this paper are deployed on a two-wheel differential mobile robot for testing in a real situation,and the experimental results are analyzed.The experimental results show that the system can locate itself in a strange indoor environment,create an environment map with high precision and good global consistency,and avoid dynamic obstacles when going to the target point.