A sensor model registration technique for mobile robot localization
S.Y.T. Lang, Andrew K. C. Wong · 2002
A method for mobile robot localization using natural environmental features is presented. An attributed graph boundary representation of free space is used. A local sensor model is registered with a global model by a subgraph isomorphism technique. Geometric and structural knowledge is used to search for mappings. The set of possible mappings is incrementally determined. Simultaneously, the associated coordinate transformation between the global and local models is refined. The model registration technique is implemented in software, and two examples are shown.>