Research On RGB-D SLAM Of Indoor Wheeled Robot

Indoor mobile robot Simultaneously Localization and Mapping(SLAM)is the basis for robot autonomous navigation and also a prerequisite for realizing intelligent robotics.In recent years,with the development of visual sensors,depth-camera based RGB-D SLAM has become the mainstream of visual SLAM research.However,the traditional RGB-D SLAM algorithm has limited information for feature matching in pose estimation,which makes the pose estimation inaccurate and sometimes leads to pose loss.In addition,due to the accumulated error,there is a drift of the estimated camera trajectory,making it difficult to obtain a globally consistent camera motion trajectory.In order to solve the problem of pose estimation,this paper presents a pose estimation method based on the combination of local map and RANSACPnP.To overcome the limitation of feature points when matching two adjacent frames in traditional RBG-D SLAM,this paper uses local map to preserve the feature point information of the previous frames as much as possible,so as to increase the number of matched feature points.In addition,the combination of RANSAC algorithm and PnP algorithm can effectively eliminate the impact of outliers on pose estimation,and obtain a more accurate pose estimation.Aiming at the accumulated error generated during the operation of the algorithm,this paper presents a pose optimization method based on local BA and global pose graph optimization.After obtaining the pose estimation,the pose optimization problem is transformed into the minimization of the reprojection error problem by the local BA,and the pose of the camera is finely adjusted to be closer to the true pose.In addition,key frames are introduced and key-frames based loop detection is performed.The keyframe pose and loop detection results are added to the pose graph for pose optimization,and globally consistent camera motion trajectories are obtained.The experimental results on the TUM dataset show that the proposed algorithm has good real-time performance and high accuracy,and can run effectively under the common indoor scenarios.