Research Of Mobile Robot Navigation Algorithm Based On Visual SLAM

Mobile robots have a wide range of needs and applications in areas such as industrial production,life services,disaster relief,and environmental detection.Navigation problems are one of the research hotspots in the field of technology.With the development of computer vision technology,the simultaneous localization and mapping(SLAM)based on vision sensor has also been deeply studied.At present,the research on visual SLAM mainly focuses on trajectory optimization and 3D environment reconstruction.It is relatively rare to apply it to mobile robot navigation.This paper takes the mobile robot navigation as the application background,and mainly analyzes and studies the navigation system design,visual SLAM algorithm and path planning.The main research work of the dissertation is as follows:(1)A visual SLAM navigation method based on graph optimization is proposed.The navigation method uses an RGB-D camera as the visual sensor,and uses the feature matching method to implement the motion estimation and loop detection of the SLAM front end,and optimizes the pose and the map point in the SLAM back end through a graph optimization method.An octree map and a two-dimensional projected grid map suitable for mobile robot navigation are constructed using the optimized pose diagrams,and a mobile robot is obtained by combining a global path planning method with a local path planning method based on the grid map.Globally optimal collision-free trajectory for navigation.(2)An RGB-D SLAM algorithm based on feature point matching is proposed.Firstly,the image feature is extracted uniformly by using quadtree-based image segmentation method,and then outlier point is excluded by the method based on k nearest neighbor(KNN)and progressive sample consensus(PROSAC)method.At last,the keyframe is used to construct the loopback,and the pose optimization is performed.The optimized point-profile diagram of the keyframe is used to construct the environment point cloud map and the navigation map.(3)A mobile robot path planning algorithm based on grid map is proposed.Firstly,for the global path planning problem,the local potential field-A*(LPF-A*)algorithm is used to detect the partial obstacle potential field of the grid map model.Then,the local potential field function of the grid node is added on the basis of the cost function of the traditional A* algorithm,and the key nodes are extracted by direction consistency to obtain the local target of the global path.Finally,the adaptive parameter dynamic window approach(APDWA)is used to implement the local path planning.Towards the objective function as the key point in the global path,and the parameters of the orientation angle evaluation function are adaptively set to meet the global conditions.Improve search efficiency with minimum offset trajectory.