Loss Tomography in Wireless Sensor Network Using Gibbs Sampling

Abstract

Listen

Translate article

The internal link performance inference has become an increasingly important issue in operating and evaluating a sensor network. Since it is usually impractical to directly monitor each node or link in the wireless sensor network, we consider the problem of inferring the internal link loss characteristics from passive end-to-end measurement in this paper. Specifically, the link loss performance inference based on the data aggregation is considered. Under the assumptions that the link losses are mutually independent, we formulate the problem of link loss estimation as a Bayesian inference problem and propose a Markov Chain Monte Carlo algorithm to solve it. Through the simulation, we can safely reach the conclusion that the internal link loss rate can be inferred accurately, comparable to the sampled internal link loss rate, and the simulation also shows that the proposed algorithm scales well according to the sensor network size.