Research On Visual Inertial SLAM Algorithms For Indoor Mobile Robot

With the rapid development of computer technology and sensor technology,humans are increasingly demanding the degree of intelligence of robots,and the use of mobile robots to perceive unknown environments is a hotspot and difficult point in robotic intelligence research.Simultaneous Localization and Mapping(SLAM)of mobile robots is one of the important basic problems.Effectively solving SLAM problems is considered to be one of the key technologies to truly make autonomous control and intelligence of mobile robots come true.The main application scenario of SLAM is that in a strange environment,the robot initially does not have information about the surrounding environment.At this time,the robot needs to receive information about the surrounding environment and its own state through sensors installed on the body,and then analyze and process the information to make the map of strange environment and know the location of your own map.This article aims at the shortcomings of the traditional visual SLAM algorithm,and builds a set of mobile robot positioning system based on visual inertia SLAM.The main idea of this paper is as follows:1.Research on visual SLAM algorithm based on improved closed-loop detection.In order to solve the problems of existing mobile robots for simultaneous localization and map construction?complicated calculation together with low efficiency of closed-loop detection,a visual SLAM algorithm based on improved closed-loop detection was proposed.The entire algorithm is divided into three threads:front-end visual odometer,back-end map construction and optimization,and closed-loop detection.In the front-end processing,the ORB algorithm is used to extract the feature points in the image,and compute robust descriptors.Aiming at the problems that the traditional RANSAC algorithm has inaccurate iterations and poor robustness when performing mismatch rejection,this article uses the PRO SAC algorithm to reject mismatch,by minimizing the reprojection error between two frames to calculate the corresponding Transformation matrix to get the pose of the robot.When the estimated pose transformation exceeds the set distance or angle threshold,the frame is taken as the key frame;in the back-end optimization,the wolf pack optimization algorithm is used to search the library for the global optimal fitness value with the current frame image,By searching for similar robust descriptors for feature matching to determine whether a closed loop is reached,and using g2o to globally optimize the pose map,thereby reducing the cumulative error,and obtaining a globally consistent pose and point cloud map.Simulation results show that the algorithm effectively improves the positioning accuracy and the efficiency of closed-loop detection,and reduces the cumulative error,while ensuring the reliability of the system.2.Research on improved visual inertia SLAM algorithm.The single feature point method only extracts the obvious feature points in the environment;however,it has poor adaptability to low-texture environments.The single direct method own higher robust in low-texture images,requires less computation,and faster speed,but it completely use the gradient search to calculate the pose of camera and easily fall into a local optimum.In practical applications,purely visual SLAM algorithms receive many restrictions,such as close obstacles blocking most of the scene,unstructured surfaces lacking visual cues,repeated textures complicate the search for corresponding relationships etc.Meantime,IMU can Brings short-term precise constraints.According to the advantages of feature point method,direct method and IMU,an improved visual inertia SLAM algorithm is proposed.Simulation results show that the algorithm effectively reduces the errors of pose and improves system accuracy in complex environments.3.The improved visual inertia SLAM algorithm is experimentally verified.Turtlebot2 mobile robot chassis and depth camera Kinect2 and inertial sensor 9Dof razor IMU were used to set up an experimental platform under the ROS development environment.The experimental platform was used to verify the effectiveness of the algorithm proposed in the paper.Figure[35]table[6]reference[61]...