Stochastic Binary Sensor Networks for Noisy Environments

This paper proposes a stochastic framework for detecting anomalies or gathering interesting events in a noisy environment using a sensor network consisting of binary sensors. A binary sensor is an extremely coarse sensor, capable of measuring data to only 1-bit accuracy. Our proposed stochastic framework employs a large number of cheap binary sensors operating in a noisy environment, yet collaboratively they are able to obtain accurate measurements. The main contributions of this paper are: (a) The theoretical accuracy analysis of the proposed stochastic binary sensor network, (b) an adaptive data collection framework based on the current measurements in order to reduce the energy consumption, and (c) a novel coding scheme for energy-efficient routing. To quantify the performance of our proposed stochastic approach, we present the simulation results of two stochastic binary sensor networks for anomaly detection using our proposed coding scheme and adaptive data gathering framework. For many scenarios, our proposed framework can reduce the energy consumption over the traditional approach by an order of magnitude.