Doctor ; The environmental modeling and localization are essential for the autonomous mobile robot system. A mobile robot requires a representation of environment which can be recognized by robotic sensors, and it should recognize its own location in the environmental model to perform autonomous navigation in the environment. Among various types of map representations and localization methods, the topological approach has an advantage of handling a large amount of data because of compact and abstracted form of environmental modeling. Moreover, it is suitable to implement a human robot interaction because the methodology of representation is similar to human. This thesis presents a method of autonomous topological modeling and localization in a home environment using only low-cost sonar sensors. In home environment, nodes should be segmented as spaces such as rooms, and only few narrow regions such as doors can be considered as edge regions. The topological localization should be performed by classifying the nodes where the robot is currently located. For this purpose, we propose an efficient method which divides whole environment into several subregions to extract topological model, and a topological localization method based on grid-map matching is proposed. Furthermore, the localization method is expanded to consider the kidnap situation of the mobile robot during operation. First, the topological model is extracted from a grid map using cell decomposition and normalized graph cut. The autonomous topological modeling involves the incremental extraction of a subregion without predefining the number of subregions. Second, a method of topological localization based on this topological model is proposed wherein a current local grid map is compared with the original grid map. The localization is accomplished by obtaining a node probability from a relative motion model and rotational invariant grid-map matching. Last, the topological localization method is modified by adding the relocation stage to solve the kidnap recovery problem. The relocation method detects the kidnap automatically and recovers it using multiple hypothesis tracking. After kidnap recovery, it also provides a criterion for selecting a reasonable hypothesis for returning to the pose tracking stage autonomously. The proposed method extracts a well-structured topological model of the environment, and the localization provides reliable node probability even when the robot is kidnapped during operation. Experimental results demonstrate the performance of the proposed topological modeling and localization in a real home environment.


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    초음파 격자지도를 이용한 위상학적 지도작성 및 위치인식 기법 ; Topological Modeling and Localization using Sonar Grid map



    Medientyp :

    Hochschulschrift


    Format :

    Elektronische Ressource


    Sprache :

    Englisch


    Klassifikation :

    DDC:    629