Talk by Mohammadreza Mahmudimanesh on 05.02.2014 at 15:00h

Title: Handling Node and Link Failures in Compressed Sensing for Wireless Sensor Networks

Abstract:

Compressed sensing is increasingly being used as an efficient network coding technique for data compression in wireless sensor networks. It is also robust against additive noise that is often present in the communication channel. While the effect of additive noise is well studied in compressed sensing, there is little or no work regarding other sources of failures that affects its performance. In particular, node and link failures are one of the most common sources of errors in a wireless sensor network. In this paper, we consider scenarios when sensor nodes or communication links encounter failures due to either external or internal causes. We show that conventional compressed sensing is unable to handle such failures. Consequently, we propose a novel and efficient technique to detect the location of failures and minimize the effect of faulty or missing sensor readings on the accuracy of the genuine data. Our fault-detection technique can detect both transient and burst errors without the overhead of exchanging health-monitoring messages.