A unified formalism for landmark based representation of maps and navigation plans

Abstract

We present a unified formalism for representing maps and using them for constructing plans of navigation for an autonomous agent. The foundation of this work lies in addressing key questions that an agent is confronted with when navigating. That is, besides the main task of how to reach the intended destination from the current position, the agent faces other questions like: where am I? what landmarks can I see? where is my destination relative to me and the landmarks I am seeing?

Fundamental to this representation is the use of visual landmarks, which are used as pivotal points in the landscape being described. Further, in the representation of spatial information and navigation there are three different viewpoints: first, the localized representation from the viewpoint of a sighted, mobile agent; second, the static representation seen by the map-maker; and third, the view of an external agent giving directions on the basis of his own experience/knowledge. The major contribution of this map model and the associated navigation method lies in the framework which unifies these three different points of view. This unification enables the agent to make no distinction in terms of following implicit instructions contained in a map and the directions given by external agents.