Research On Visual SLAM And Path Planning Of Mobile Robot Based On Multi-features Fusion

Visual Simultaneous Localization and Mapping technology,as a key technology to solve the problem of robot positioning and perception of the surrounding environment,has been widely used in many fields such as automatic driving,3D reconstruction and mobile robot navigation in recent years.With the emergence of depth cameras,the visual SLAM system based on depth cameras has received extensive attention due to its advantages of low cost,rich information,and portability.In addition,path planning technology,as a key technology for autonomous movement of mobile robots,also has important research value.This paper mainly focuses on the RGBDSLAM and path planning technology based on multi-feature fusion.The work includes:(1)In order to solve the problem of low positioning accuracy of the traditional point-based SLAM algorithm in weak texture scenes,a RGB-DSLAM method based on multi-feature fusion is proposed by fusing point-line-surface features.In order to improve the uniformity and extraction speed of the feature point extraction module,the SLAM system in this paper adopts the improved AGAST algorithm based on quadtree as the system point feature extraction algorithm.The line feature and surface feature extraction modules use EDLines and hierarchical clustering algorithms to improve the accuracy and speed of feature extraction.In addition,this paper utilizes the geometric properties of line features to improve its matching speed during line feature matching.At the same time,in order to improve the matching accuracy of surface features,a combination of conventional matching methods and similarity measures is used for surface feature matching.Finally,in order to improve the positioning accuracy of the system,the virtual right-eye line feature constraints and hypothetical plane constraints are introduced to construct the total reprojection error function,and the factor graph is constructed to optimize the pose.(2)In order to realize real-time obstacle avoidance while planning the optimal path for mobile robots in complex dynamic environments,a path planning algorithm integrated with improved DWA is proposed.Aiming at the problem that the DWA local path planning algorithm is easy to fall into the local optimum in the scene with many obstacles,an intermediate target point selection strategy is introduced,and the nodes that meet the requirements are selected from the path node set as DWA through this intermediate target point selection strategy.The intermediate target point of,which is used to guide the local path planning.Then the algorithm is fused with the improved DWA algorithm,and the simulation experiment is carried out through MATLAB.The experiment proves that the fusion algorithm in this paper can plan an optimal global path in a more complex environment while dynamically avoiding obstacles in real time.(3)Using the TUM dataset,compare the positioning accuracy of the RGBDSLAM system based on multi-feature fusion with ORB-SLAM2 and other open source VSLAM frameworks.Experimental results show that the SLAM system proposed in this paper has better positioning accuracy and performance.In addition,in order to further verify the mapping ability of the SLAM system in this paper,the threedimensional dense point cloud map and octree map of the environment were constructed through the dataset,and real-time mapping experiments were carried out on the mobile robot.Finally,the fusion path planning algorithm in this paper is applied to the mobile robot for experiments.The experiments show that the fusion path planning algorithm proposed in this paper can realize real-time obstacle avoidance while planning the optimal path for mobile robots in complex environments.