Research And Application Of Mobile Robot Based On Visual SLAM

In the field of robotics,visual SLAM(Simultaneous Localization and Mapping)technology has become one of the most discussed topics in the robotics industry.SLAM technology is an autonomous positioning and mapping technology for robots in an environment without prior conditions.With the development of science and technology,the idea of using cameras to realize spatial three-dimensional reconstruction is gradually being promoted and applied.This paper mainly studies the improved algorithm based on ORB-SLAM(Oriented and Brief-SLAM)technology.It starts from the three aspects of front-end,back-end and mapping technology,improves the existing visual SLAM algorithm,and applies it to actual mobile robots.In the front-end visual odometer,the SIFT algorithm is used to detect the extreme values of the Gaussian pyramid in the detection of key points to improve the key point detection technology of the original ORB algorithm,which makes the ORB key points more robust and does not affect real-time performance.In the back end,in order to improve the efficiency and quality of the back end optimization,a BA optimization model is constructed,and the least squares error term model is used as the cost function.The Jacobian matrix sparsity of the incremental equation is used to optimize the real-time performance of the BA.At the same time,for large-scale maps and path structures,the simplified pose map Pose Graph is used to abandon repeated optimization of a large number of feature point landmarks in the original BA optimization,which reduces the BA computation load.In the construction map,the PCL point cloud map occupied too much memory.Using Octomap mapping technology,the monocular RGB-D camera was used to obtain the depth map and color map,and the path and Octomap map obtained by the ORB-SLAM algorithm.Correspondingly,a 3D-RRT-ISM map construction technology is proposed.In the experiment,it is divided into the optimization of the robot's motion path and the obstacle avoidance.The laser sensor is used as the auxiliary to achieve the robot's obstacle avoidance and SLAM mapping using the depth camera.In terms of path planning,RRT fast random tree is used to obtain obstacle avoidance blank area nodes in the map.Subsequently,the improved ant colony algorithm was optimized by optimizing the heuristic function to navigate the robot's walking path from the shortest time and shortest distance.The terminal uses the visual interface of the robot host computer designed and developed by Qt to display the SLAM map and Robot motion path.