Adaptive Data Collection in Environmental Monitoring Networks

Outline:

Power-budgeting is one of the fundamental challenges in wireless sensor networks today. As the size of typical environmental monitoring networks grow, the severity of the power problem gets magnified. Adaptive collection of data aims at reducing the power challenges by reducing the amount of data transfered from the mote to the base-station, and by reducing sensing and storage costs. However, we note that an effective adaptive data collection scheme requires a deeper understanding of the spatial and temporal variation exhibited by the modalities of interest. Towards this goal, we are exploring a principal component analysis (PCA) based method to study and understand the heterogeneity in typical environmental monitoring networks. This analysis will enable us to design a system that collects data adaptively so that the power in utilized more effectively.

Figure shows an illustration of the sharp transients shown by air temperature due to precipitation. One notices the deviation of the air temperature signal from the expected diurnal pattern. Our previous work has demonstrated that a PCA-based method is very effective to characterizing and detecting such informative periods. We are extending this work to effectively characterize the spatial heterogeneity, in addition to temporal heterogeneity.

Publications:

• Jayant Gupchup, Andreas Terzis, Randal Burns and Alex Szalay. Model-Based Event Detection in Wireless Sensor Networks. Appeared in the Proceedings of the Workshop for Data Sharing and Interoperability on the World Wide Web (DSI 2007). April 2007. PDF