Research On Indoor Environment Map Building Based On Pose Graph And Placement Identification Method

Autonomous mobile robot has become the important direction of robot technology in our country. Especially in the face of the aging population, service robot industry will have a period of rapid development.The key technology of service robot includes positioning technology, navigation technology and interaction technology, etc. The stability and safety of the service robot have become its development bottleneck. Considering long-term work of service robot in indoor environment, this paper mainly researches on simultaneous localization and map building and environment recognition technology. The main contents are as fllows:Service mobile robot prototype is employed as research object, whose chassis adopts differential drive way, equipped with laser range sensor. Mobile robot system model was established, including the differential drive model, odometer model and laser sensor measurement model as the foundation for map building.Service robot need interact with environment for a long time. To solve the adaptability problem of long-term map building, SLAM method based on pose graph was used for map building. Bayesian reliability network was employed to describe the SLAM problem of robot. Sparse matrix decomposition and updating were adopted to solve the robot pose and landmark. In addition, matrix and graph model were combined to express the process of the variable elimination algorithm. Maximum likelihood method was adopted for data association, and mahalanobis distance was used to consider uncertainty between robot pose and landmark instead of Euclidean distance.A method was proposed for placement identification based on indoor structured environment to solve the problem of robot “kidnapping” and topology information access. This paper put forward a corridor identification method based on the direction variance distribution of laser sensor data through SVM. After data preprocessing, the direction variances were calculated to describe environment information. A SVM was applied to make corridor identification.And for indoor environment identification between each room, a method based on the optimal path was proposed considering the layout of each room. Sample points were set in all the room and the distance between sample points was calculated to obtain the probability of observation. Then, the possible optimal paths under the time sequence were obtained according to the HMM model, and through the D-S reasoning algorithm, room identification method was obtained to complete indoor environment topology placement access.Finally, experiments was carried out on SLAM, corridor recognition and room identification in Robotics Institute to verify the validty of the overall research.