Component Based Clustering in Wireless Sensor Networks

TL;DR

This paper presents a component-based architecture for clustering algorithms in wireless sensor networks, enabling easy comparison, reuse, and scalability of well-known algorithms across different hardware and software platforms.

Contribution

It introduces a modular framework for developing and evaluating clustering algorithms, facilitating rapid prototyping and cross-layer integration in WSNs.

Findings

Implemented 5 algorithms and discussed 11 more for the framework.

Simulation results show high fidelity to original algorithms and scalability.

Practical experiments demonstrate applicability in real WSNs.

Abstract

Clustering is an important research topic for wireless sensor networks (WSNs). A large variety of approaches has been presented focusing on different performance metrics. Even though all of them have many practical applications, an extremely limited number of software implementations is available to the research community. Furthermore, these very few techniques are implemented for specific WSN systems or are integrated in complex applications. Thus it is very difficult to comparatively study their performance and almost impossible to reuse them in future applications under a different scope. In this work we study a large body of well established algorithms. We identify their main building blocks and propose a component-based architecture for developing clustering algorithms that (a) promotes exchangeability of algorithms thus enabling the fast prototyping of new approaches, (b) allows cross-layer implementations to realize complex applications, (c) offers a common platform to comparatively study the performance of different approaches, (d) is hardware and OS independent. We implement 5 well known algorithms and discuss how to implement 11 more. We conduct an extended simulation study to demonstrate the faithfulness of our implementations when compared to the original implementations. Our simulations are at very large scale thus also demonstrating the scalability of the original algorithms beyond their original presentations. We also conduct experiments to assess their practicality in real WSNs. We demonstrate how the implemented clustering algorithms can be combined with routing and group key establishment algorithms to construct WSN applications. Our study clearly demonstrates the applicability of our approach and the benefits it offers to both research & development communities.