Multivariate context collection in mobile sensor networks

We focus on the treatment of quality-stamped contextual information in mobile sensor networks. Sensing nodes capture and forward context for consumption by mobile context aware applications. Due to the dynamic network topology the quality indicators seen by consumers vary over time. Context quality is a decreasing function of time and context can be consumed with a certain delay from its capturing time. We propose the sequential assessment of the network-circulated context information according to the Generalized Secretary Problem, a known paradigm in the Optimal Stopping Theory. The consumer node delays the processing (consumption) of incoming context until better quality is attained. We extend this basic model to include the cardinality of contextual components (i.e., different types of measurements coming from, possibly, different sources). Hence, the consumer node is interested not only in the higher possible quality of context but also in the widest possible range of context parameters (context “quantity”). We compare our findings to simple consumption strategies and pinpoint the advantages of the proposed model