Quality-aware Service Selection Approach for Adaptive Context Recognition in IoT
While developing context-aware applications, there may be uncertainty with respect to the available data sources. Applications that are developed to a fixed set of data sources may not be flexible enough, to adapt to change in the sensing environment such as sensor disappearance or degradation. Opportunistic sensing tackles this problem by enabling the automatic detection and selection of data sources. However, existing approaches rely on a limited number of quality metrics and do not take into account the influence of data processing on the quality of context recognition. In this paper, we present an extension of the opportunistic sensing approach that is able to take into account quality metrics like execution time affecting the overall quality. Our approach consists of modelling of available data sources and data processing methods that can be used to assemble context recognition chains and estimate their quality. We present a prototypical implementation of the models and mechanisms in an autonomous driving test environment and provided testing results on a use case for finding traffic congestions in a specified route.