QoINF: Quality-of-Inference(QoINF)-Aware Context Determination
Two main observations further drive our work. First, the accuracy of inferred context increases with the use of a progressively larger sensor set. Pervasive computing environments typically contain many sensors capable of sensing the same context to different degrees of accuracy. Second, there is effectively a tradeoff between the energy overheads of monitoring and the achievable QoINF value. The quality of the inferred context is a function not just of the chosen sensors but also of the permitted inaccuracy in the sensed values. We derive a generalized function that relates an application's specified required QoINF value to the set of sensors used and their associated allowable imprecisions. We demonstrate how this function may be computed from observed empirical data and use experimental sensor data to quantify our selection heuristic's accuracy and its reduction in overhead.
The integration of sensor networks into pervasive computing is enabling detailed monitoring of the network, the environment, and individuals in the environment; this increase in sensing capability opens the question of how to efficiently and reliably convert sensed data into high-level abstractions of context. The determination of a specific context attribute may be viewed as an inference obtained by fusing the values from multiple sensor streams. Our central concept to this end is a Quality-of-Inference (QoINF) specification, defined as the error probability in estimating a context state, given the imprecision in the values of the contributing sensors.
QoINF-Aware Context Determination Architecture
Papers
- Resolving and Mediating Ambiguous Contexts for Pervasive Care Environments
- MobiQuitous 2009, Poster Paper
- Resource-Optimized Quality -Assured Ambiguous Context Mediation in Pervasive Environments
- QShine 2009
- A Prototype for Resource Optimized Context Determination in Pervasive Care Environments
- MobiCASE 2009, Demonstrations Track
- Design and Evaluation of a Model based Quality Sensitive Context Determination Framework
- Under Review
