Research And Implementation Of VSLAM Technology For Mobile Robot In Dynamic Environment

With the speedy development of science and technology,technologies associated with robots have gradually entered people's lives.Simultaneous localization and mapping(SLAM)algorithm is one of the most important technologies of the robot to realize autonomy,algorithms based on vision sensors have become a hot research of intelligent robotics due to the inexpensive and wealthy image information of the camera.In this thesis,the research status of SLAM technology at home and abroad is elaborated,and focusing on the research of visual SLAM related algorithms that can be used in dynamic environments.After analyzing the related technologies of the robot navigation system and sensors,the RGB-D camera that can directly obtain image depth was chosen as the vision sensor of the navigation system.Model the robot system to build a mobile robot VSLAM system that can run in a dynamic environment.In the research of visual SLAM,In view of the low robustness and poor accuracy of traditional visual odometry in dynamic environment,a dense visual odometry algorithm based on edge information fusion was proposed.By adding edge information,the algorithm performs edge extraction and clustering according to the current frame to calculate the residual model of the fusion DT error.The average background depth is used to enlarge the residual gap between the dynamic and static clusters to assist motion segmentation,and the edge alignment error is added as a constraint to the pose estimation to calculate a more accurate estimation result.The added edge information makes the algorithm more robust in a variety of different environments.Experiments on TUM dataset show that the proposed algorithm had higher accuracy and robustness than the existing algorithm in dynamic environment,it can achieve better results in a static environment.In the research of path planning problem,an improved Informed RRT* path planning algorithm is proposed to solve the problem that the path searched by the traditional RRT algorithm has many corners and the path is not the shortest.The improved algorithm adopts the strategy of goal-oriented and variable step length to increase the probability of the algorithm growing towards the goal and improve the search efficiency.In addition,a cubic B-spline curve is used to optimize the searched path to remove the corners of the path,and obtain a smoother and lower cost optimal path.The experiment results show that the improved algorithm has better comprehensive performance than other RRT path planning algorithms.Finally,the construction of the mobile robot VSLAM navigation system is completed,and experiment with the improved visual odometry algorithm and path planning algorithm on the system in the actual dynamic environment,the functions of positioning,mapping and navigation of the mobile robot are realized.The experimental results show that the improved algorithm has better performance in the actual dynamic environment and can run stably on the mobile robot system.