Research On Indoor Mapping And Path Planning Method Based On Visual SLAM Technology

With the development of computer technology,vision-based Simultaneous Location and Mapping(SLAM)has become a key technology in intelligent indoor mobile robots,and the use of low-cost RGB-D sensors for indoor mobile robot environment sensing has become a popular research direction.Although it has been studied for many years,it still faces many challenges.In this regard,this thesis designs a visual SLAM system of " Indoor Mapping + Path Planning " through in-depth research on indoor mapping and path planning method based on visual SLAM technology,and verifies its effectiveness through experiments.The specific work of this thesis is as follows:(1)To overcome the problem of inaccurate camera pose estimation,this thesis proposes a joint pose optimization method based on sparse feature points and dense pixel points.The method uses the scale-invariant feature transform(SIFT)feature point descriptor to match sparse feature points between frame blocks,and then calculates the Euclidean distance of matched point pairs,and combines the luminance difference and plane distance of dense pixel points between frame blocks to optimize the camera pose in blocks by Gaussian Newton method.The experimental results show that the joint positional optimization method proposed in this thesis outperforms the iterative closest point(ICP)algorithm in terms of overall stability and accuracy.(2)To solve the problems of information redundancy and map ghosting in the globally consistent point cloud model.In this thesis,we propose a point cloud fusion method based on truncated signed distance function(TSDF).This method adopts the truncated signed distance function to improve the quality of indoor reconstructed scenes by weighted fusion and weighted separation of point clouds,and overcome the situation that indoor point clouds have ghosting caused by wrong bit pose.The feasibility of the proposed method is verified on the STANFORD indoor dataset.(3)The traditional path search algorithm cannot meet the requirements of safety and kinematics feasibility in the traveling process of the mobile car,and this thesis proposes a path planning method of mobile car based on Uniform B-splines curve.The front-end adopts a kinodynamic path searching method to find a collision-free initial path in the discreted control space.The back-end uses the convex hull properties of the uniform B-spline curve to effectively combine the distance information of obstacles and the motion constraints of moving carts to optimize the smoothness and safety of the initial path by the soft constraint method.Finally,the re-planning path mechanism is adopted to enable the mobile car to find a safe path in a dynamic environment.The experimental results show that the method has good safety and kinodynamic feasibility in various complex simulation environments.