Robust and distributed localization in sensor networks

We present a robust localization system allowing Wireless Sensor Networks (WSNETs) to determine the physical location of their nodes. The coverage area is split into partitions and the system seeks to identify the partition where a sensor resides based on observations made by stationary sensors (clusterheads). These observations are assumed random and are often RF-characteristics of the signal transmitted by the sensor and received by the clusterheads. To every partition-clusterhead pair we associate a family of probability density functions (pdfs) from which observations are assumed to be drawn and pose the localization problem as a composite multi- hypothesis testing problem. We establish conditions under which the Generalized Likelihood Ratio Test (GLRT) is optimal in a Generalized Neyman-Pearson (GNP) sense. We obtain the optimal GLRT threshold to minimize the maximum probability of error and establish a probabilistic performance guarantee for the resulting decision rule. We study the problem of optimally placing clusterheads to minimize the maximum probability of error and establish a performance guarantee for the optimal placement. We also demonstrate how the localization system can be implemented in a distributed manner by appropriate in-network processing. The proposed approach is validated in a testbed yielding promising results.