Research On Visual Simultaneous Localization And Mapping Of Miniature Mobile Robot

Simultaneous localization and mapping(SLAM)has always been a valuable research issue in the field of robotic automation.It can realize the autonomous positioning and the establishment of the environment map according to the pose estimation during the robot movement and the environmental information acquired by sensors.Therefore,this paper focuses on SLAM technology for mobile robots in unknown environments.Firstly,the background and significance of SLAM technology are introduced,and the research statuses both domestic and international are described.The related knowledges of visual SLAM technology are elaborated,including SLAM problem description,vision sensor,Kalman Filter and Extended Kalman Filter.The design process of motion model,vision model,landmark observation model and noise model of vision sensor are introduced emphatically.At the same time,the filtering principles of Kalman Filter and Extended Kalman Filter are described.Secondly,an improved algorithm for feature point detection is proposed.This algorithm adds pre-detection steps on the basis of FAST algorithm,and simplifies the feature descriptor of SIFT algorithm.Through simulation and contrast experiments,it is proved that the improved algorithm can effectively improve the detection of visual landmark information.Thirdly,for the data association problems,a data association algorithm based on unilateral and bidirectional matching is proposed.The algorithm performs the unilateral matching according to Euclidean distance to delete feature points that do not satisfy the condition,and performs the bidirectional matching according to Mahalanobis distance.The image feature point matching simulation experiment is carried out by using this algorithm,and the superiority of the algorithm is proved by comparison.Finally,the system working interface is designed.By setting the environmental roadmap and the moving trajectory of the mobile robot,the simulation experiment are carried out based on the improved algorithm and the original algorithm in EKF-vSLAM system.The simulation verifies the performance of the improved algorithm in terms of the pose estimation and the ability of reducing the error.