Knowledge Representation and Inference in Context-Aware Computing Environments

Abstract

The present document provides a comparison of different knowledge representations and their inference models that can be used for context aware computing. In order to execute inference, any ubiquitous computing environment has to maintain the existing knowledge. The way of representing this information in the knowledge representation has a great influence on the performance that the inference system would carry out. An example scenario whose purpose is to avoid rearend collision in a vehicular environment serves as basis to derive requirements for the knowledge representation and inference. We try to apply each approach to the example scenario (or parts of it) to obtain the benefits and limitations. Like that we come to the conclusion that Bayesian Networks are the way that fits best for a ubiquitous computing scenario.