Dynamic modeling of an indoor environment

Proposes to exploit an uncertain man-made representation of an indoor environment. The problem of uncertainties valuation is solved by a human behavior modeling which consists in using proximity relations between the geometrical primitives of the objects. The multivalue numbers formalism provides an efficient structure in order to model the a priori representation. The dynamic aspect of such a model makes real-time alterations of the map components possible. A Kalman filter used in order to match ultrasonic data with the preprocessed map while navigating. This process allows the map to be incrementally improved while preserving the original shape of the obstacles.