Event-driven field estimation for wireless sensor networks

Abstract

This paper introduces and analyzes a field estimation scheme for wireless sensor networks. Our scheme imitates the response of living beings to the surrounding events. The sensors define their periphery of attention based on their own readings. Readings differing from the expected behavior are considered events of interest and trigger the data transmission to the sink. The presented scheme is evaluated with real-site-collected data and the tradeoff between the amount of data sent to the sink and the reconstruction error is analyzed. Results show that signifcant reduction in the data transmission and, as a consequence, in the energy consumption of the network is achievable while keeping low the average reconstruction error.