Path Planning Of Mobile Robot Based On Visual Information In Dynamic Environment

Path planning is a necessary step for a mobile robot to perform its tasks autonomously.In the dynamic environment,that is,when there are moving obstacles in the map,how to carry out the optimal path planning is always a problem that is highly valued by academic circles at home and abroad.Therefore,based on the existing problems,the thesis designs an environment-aware algorithm to identify the dynamic obstacles.Based on this,the algorithm for obtaining the optimal path is studied.Firstly,through the binocular visual perception technology,the ground plane detection algorithm of mobile robot on flat land is studied and designed for indoor and urban highway environment.The spatial 3D point cloud information and the 2D image information captured by the camera are merged,and the feasible path detection is realized by the clustering algorithm,which effectively improves the running speed of the algorithm.Then,the visual SLAM is used to construct a map of the surrounding environment of the mobile robot and obtain its own pose in real time.The visual algorithm is integrated into the improved DRE-SLAM algorithm based on ORB-SLAM to realize the effective recognition of dynamic obstacles.And then the obstacles are removed,leaving a static background for effective map construction.In the detection of the target,the deep learning algorithms YOLOv3 and Mask R-CNN are combined to adapt to different environments,and the static background pixels are eliminated.After the SLAM algorithm completes building a map in a dynamic environment,it can determine the state of obstacles.The path planning algorithm in a static environment can be extended to a dynamic path planning algorithm by using this information.Therefore,the thesis first designs an optimal path planning algorithm where potential function is applied to the construction region in a static environment,and obtains the initial solution.After obtaining the initial solution,the dynamic obstruction is detected by constructing an elliptical exclusive space.The range of the exclusive area is determined by the value of the mobile robot’s operating speed,and its purpose is to reduce the required detection range.When the dynamic obstacle enters the exclusive area,it simulates human decision-making and uses fuzzy control algorithm to realize local optimal path planning.After successfully avoiding obstacles,it reconnected to the initial solution of static path planning,and realized the global optimal path planning.On the basis of theoretical research,the ROS operating system is used to fuse the feasible path algorithm,SLAM algorithm and visual detection information to construct the map and realize the positioning of the mobile robot.The path planning algorithm is integrated into ROS through plug-ins.In order to verify the feasibility of the algorithm designed in this paper,the KITTI data set was selected for testing,especially for people and cars moving in outdoor scenes.The system simulation was carried out with the help of Gazebo,and the path information of the mobile robot planning was displayed in real time through the rviz tool in ROS.