Markov Random Fields based probabilistic relaying for sensor networks

In this paper, we demonstrate how a Markov Random Field (MRF) based framework can be used for sensor networks analysis and design. In fields such as image processing it has been shown that MRFs is a powerful tool to analyse distributed systems with strong spacial interactions, which is also a defining characteristic of sensor networks. In this work we focus on using MRFs to model traffic intensity of sensor networks using shortest path routing. Later we propose a probabilistic relaying mechanism to recreate a traffic pattern similar to that observed in a network using shortest path routing. The objective is to emulate the shortest path performance without complex routing protocols and associated overheads. Using a simulation study we then show that the proposed mechanism achieves 95% of the throughput of shortest path, without using a routing protocol.