ABSTRACT

We aim at developing software agents that are able to simulate the mechanisms involved when manipulating spatial mental images and reasoning about them. In order to simulate the way people usually reason about space, we need a model of space that allows qualitative reasoning. The Voronoï model of space is a good candidate for such a requirement. It provides an unambiguous way of defining neighbors based on relative proximity, and of constructing topological networks that encapsulate the spatial relationships between objects. In this chapter we show how various spatial prepositions can be mapped onto topological configurations in the Voronoï model. We show how we can develop a Voronoï agent that has the responsibility for maintaining a spatial database, for interpreting queries and commands expressed with prepositional descriptors, and for reasoning about the spatial relationships that result from these interpretations. In a multiagent system, the Voronoï agents are the only agents that handle spatial information. Other agents send queries to the Voronoï agents in order to get any topological or metric information characterizing the spatial relationships of the objects that are part of the space region covered by the Voronoï model. The Voronoï agents and their databases simulate a cognitive map whose information can be accessed by other agents. In addition, the information contained in these cognitive maps can be displayed to the user using familiar cartographic conventions.