A statistical model for fault inference in Wireless Sensor Networks
LE3 .A278 2008
2008
Shakshuki, Elhadi Zhang, Ying
Acadia University
Master of Science
Masters
Computer Science
An embedded sensor network is a system of nodes, each equipped with a certain amount of sensing, actuating, computation, communication and storage resources. Two major components of sensor nodes are a sensing unit and a wireless transceiver. They interact directly with nodes in Wireless Sensor Networks (WSNs) that are easily prone to failure due to hardware failure, communication link errors, energy depletion, malicious attack, etc. Even if the sensor node hardware is good, still the communication among sensor nodes depends on many factors such as signal strength, obstacles, and interference. Degradation of these factors results in low reliability of performance of sensor nodes. One of the key prerequisites for an effective and efficient embedded sensor system is the development of a low cost, low overhead, highly resilient fault-inference technique. To address the issue of fault inference in micro-sensors, a statistical model is proposed. The proposed model involves a fault inference engine. This engine profits from an Expectation Maximization algorithm to evaluate sensor nodes' fault probabilities. Theoretical analysis and simulation experiments are conducted to demonstrate the effectiveness of this model.
The author retains copyright in this thesis. Any substantial copying or any other actions that exceed fair dealing or other exceptions in the Copyright Act require the permission of the author.
https://scholar.acadiau.ca/islandora/object/theses:3001