GEOMETRIC UNCERTAINTIES IN POLYHEDRAL OBJECT RECOGNITION

Data gathered from robotic sensors inherently contain uncertainties or errors that cause them to deviate from their nominal values. When the data are interpreted as coming from the surface of a known object, the initial errors propagate. As a result, there are further uncertainties about which object best interprets the data and also about where in space the object may lie. It is this latter uncertainty that this paper addresses. We show, by direct geometric construction, that previous uncertainty bounds on the location of polygonal or polyhedral objects can be tightened considerably. The improvement of the bounds is a result of considering the cross-coupling between rotational and translation uncertainties in the interpretation of the sensor data.