Introduction. A key challenge in pervasive systems is that of identifying and developing appropriate formal approaches for understanding and reasoning about contexts. This is a di cult task since there is no common agreement on what constitutes a context and what its properties are. Although the literature contains a rich variety of context models [1, 2], most of the resulting implementations are domain-dependent and lack powerful reasoning capacities. Preliminary work on logic-based reasoning about contexts has been proposed in [3, 4], using bridge rules and argumentation theory. However, application-wise there are still open questions on how to model context-related pieces of knowledge in general domains, so that appropriate reasoning techniques can be applied. In order to put the basis for a comprehensive theory of contexts one should take into account the general structure of knowledge needed to represent contexts, in particular, what is intended as a context in pervasive environments and which components/properties should be identi ed and described. System properties that are of interest for the speci cation of contexts should be also considered. Motivated by application challenges, this short paper highlights observations that can be used to address these issues.
Generalization of Context in Pervasive Environments. Previous investigations
of context models in pervasive environments revealed that within contexts
certain entities are more important than others in practice. These are location,
identity, activity and time [
To enhance generality and modularity of the representation, we suggest to represent entities at a primary level while all the other pieces of sensed information are at a second level and can be indexed as attributes of the entities at primary level. As for the relations between subjects, objects, and places in pervasive environments the basic interesting relations that should be formalized are spatial (being close/far, being in/out), temporal (occurring before/after) and causal (occurring because of something).
Properties. In a given context, attributes and relations need to be associated to a time stamp. In this way, the reasoning system can take into account their dynamic evolution to characterize aspects such as commonsense, temporal reasoning, belief revision and update. In general the temporal evolution of a context is one of the most important aspect to be correctly and formally characterized. Additional properties may be needed to determine what is true in a given context where certain objects, attributes and entities have been de ned. These properties should make it possible to express belief merging and belief revision/update in contexts.
Conclusions. We believe reasoning about contexts became harder and harder because of the proliferation of very personal representations of context knowledge in terms of admissible relations, their arity and their properties, which have been developed ad-hoc for very diverse speci c applications. Our attempt is that of providing a more general and modular representation of context knowledge. The intuition is that in this way multi-context reasoning could be supported by the veri cation of formal properties at a knowledge representation level, in the same way as human deduction is based on roles of the grammatical components of a sentence.