The Role of Probabilistic Schemes in Multisensor Context-Awareness

This paper investigates the role of existing "probabilistic" schemes to reason about various everyday situations on the basis of data from multiple heterogeneous physical sensors. The schemes we discuss are fuzzy logic, hidden Markov models, Bayesian networks, and Dempster-Schafer theory of evidence. The paper also presents a conceptual architecture and identifies the suitable scheme to be employed by each component of the architecture. As a proof-of-concept, we will introduce the architecture we implemented to model various places on the basis of data from temperature, light intensity and relative humidity sensors