Dealing with application requirements and energy consumption in wireless sensor networks: a novelty detection approach for quality of query services

This paper presents an approach for adapting query processing in wireless sensor networks (WSN) based on the notions of quality of query services (QoQS) and novelty detection (ND). While the former concept captures the idea of possibly having different queries serviced in different ways by the same WSN, the second relates to a machine learning technique embedded in the WSN components that allows them to modify their query processing behavior in a dynamic fashion. This approach aims at the intelligent consumption of the limited resources available in these networks while still trying to deliver the data quality as expected by their client applications. In this context, four classes of QoQS (CoQoS) have been specified having in mind distinct levels of requirements in terms of accuracy and temporal behavior of the sensed data. Moreover, a new ND-based algorithm, named as AdaQuali (after ADAptive QUALIty control for query processing in WSN), is introduced in detail as a way to control the sensor node activities through the adjustment of their rates of data collection and transmission. For validation purposes, experiments with a simulation prototype have been conducted over real data, and the results achieved so far point to gains in terms of energy consumption reduction that vary from 1.73% to 42.99% for different CoQoS.