Locating a mobile robot using local observations and a global satellite map

Abstract

A method for locating an autonomous mobile robot using local observations and a global satellite map is described. The satellite map provides approximate elevation data for an area within which the robot is known to be located. The map consists of a coarse grid of rectangular regions annotated with upper and lower bounds on the elevation within the region. In exploring its environment, the robot makes measurements to extract information about the relative position and orientation of local landmarks. These landmarks are integrated into a stochastic map, which is then matched with the satellite map to obtain an estimate of the robot's current location. The landmarks are not explicitly represented in the satellite map. The results of the proposed matching algorithm correspond to a probabilistic assessment of whether or not the robot is located within a given region of the satellite map. By assigning a probabilistic interpretation to the information stored in the satellite map, it is possible to provide a precise characterization of the results computed by the matching algorithm.<>