Perceptual Components for Context Aware Computing James L. Crowley 1 , Joëlle Coutaz 2 , Gaeten Rey 2 and Patrick Reignier 1 1 Laboratoire GRAVIR, INRIA Rhône Alpes, 655 Ave de l’Europe, F-38330 Montbonnot, France {Crowley, Reignier }@inrialpes.fr http://www-prima.imag.fr 2 CLIPS-IMAG, BP 53, Université Joseph Fourier F-38051 St. Martin D’hères, France {Coutaz, Rey}@imag.fr Abstract. In this paper we propose an ontology and a software architecture for observing and modeling context and situation. We are especially con- cerned with the perceptual components for context awareness. We propose a model in which a user’s context is described by a set of roles and relations. Different configurations of roles and relations correspond to situations within the context. We define an ontology for context awareness from both a bottom up system’s perspective and a top-down users’ perspective. As we define each element, we describe the corresponding components of a proc- ess-based software architecture. Using these components, a context is trans- lated into a federation of observational processes. This model leads to an ar- chitecture in which reflexive elements are dynamically composed to form federations of processes for observing and predicting the situations that make up a context. 1 Introduction Available technologies increasingly enable computing and communication to migrate out of the “gray box” and into ordinary objects. An inevitable result is a multiplica- tion of digitally controlled devices with increasingly complex capabilities. In too many cases, the designers of such devices are forced to rely on the human capacity to learn. While standardization of interaction techniques may provide some help, ordinary humans are increasingly required to divert attention to “futzing with the device”. There is an obvious need for methods for building systems that model the activities a human users and anticipate their needs. Much of the work on such systems focuses on defin- ing and modeling “context” for interaction. In this paper, we propose an ontology and a software architecture for modeling con- text and situation. A key aspect of our approach is that we recognize that a context aware system must be able to sense users and their activities. Unlike much of the