Computer Science ETDs

Publication Date

12-1-2012

Abstract

This dissertation proposes an approach to creating robust communication systems in wireless sensor networks, inspired by biological and ecological systems, particularly by evolutionary game theory. In this approach, a virtual community of agents live inside the network nodes and carry out network functions. The agents use different strategies to execute their functions, and these strategies are tested and selected by playing evolutionary games. Over time, agents with the best strategies survive, while others die. The strategies and the game rules provide the network with an adaptive behavior that allows it to react to changes in environmental conditions by adapting and improving network behavior. To evaluate the viability of this approach, this dissertation also describes a micro-component framework for implementing agent-based wireless sensor network services, an evolutionary data collection protocol built using this framework, ECP, and experiments evaluating the performance of this protocol in a faulty environment. The framework addresses many of the programming challenges in writing network software for wireless sensor networks, while the protocol built using the framework provides a means of evaluating the general viability of the agent-based approach. The results of this evaluation show that an evolutionary approach to designing wireless sensor networks can improve the performance of wireless sensor network protocols in the presence of node failures. In particular, we compared the performance of ECP with a non-evolutionary rule-based variant of ECP. While the purely-evolutionary version of ECP has more routing timeouts than the rule-based approach in failure-free networks, it sends significantly fewer beacon packets and incurs statistically fewer routing timeouts in both simple fault and periodic fault scenarios.

Language

English

Keywords

Wireless Sensor Networks, Agent-based systems

Document Type

Dissertation

Degree Name

Computer Science

Level of Degree

Doctoral

Department Name

Department of Computer Science

First Advisor

Bridges, Patrick

First Committee Member (Chair)

Moses, Melanie

Second Committee Member

Ackley, David

Third Committee Member

Caudell, Thomas

Share

COinS