Design and Implementation of a Robust Sensor Data Fusion System for Unknown Signals

In this work, we present a robust sensor fusion system for exploratory data collection, exploiting the spatial redundancy in sensor networks. Unlike prior work, our system design criteria considers a heterogeneous correlated noise model and packet loss, but no prior knowledge of signal characteristics. The former two assumptions are both common signal degradation sources in sensor networks, while the latter allows exploratory data collection of unknown signals. Through both a numerical example and an experimental study on a large military site, we show that our proposed system reduces the noise in an unknown signal by 58.2% better than a comparable algorithm. Key words: Exploratory Data Collection, Robust Distributed Sensing, Data Fusion

• Type: Conference
• Year: 2010
• Where: DCOSS
• Authors: Younghun Kim, Thomas Schmid, Mani B. Srivastava
• Keywords: DCOSS 2010, Distributed And Parallel Computing, Exploratory Data Collection, Sensor Networks, Unknown Signal