Title

Swarm intelligence based methods for decentralized sensor networks

Date of Award

2009

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Electrical Engineering and Computer Science

Advisor(s)

Lisa Osadciw

Keywords

Swarm intelligence, Sensor networks, Distributed networks, Decentralized networks

Subject Categories

Electrical and Computer Engineering | Engineering

Abstract

This dissertation proposes an adaptive autonomous sensor manager for a distributed, decentralized sensor network. Swarm intelligence based algorithms are designed to control, optimize and organize decentralized sensor networks. The concepts of universality, automatic, adaptivity and unrestrictive, algorithm design are defined in the context of sensor networks. Examples are given showing the efficacy of swarm intelligence algorithms in each of these contexts.

Swarm intelligence is a bottoms up approach for control and optimization problems. Individual interactions among agents result in an emergent behavior that is utilized to solve the problems. The emergent behavior can be used to optimize a performance function (centralized), organize the networks reacting to real time changes (decentralized) or simply control the decision making. The successful use of emergent behavior involves design of probabilistic rules for agent decision making, and deterministic rules for agents interaction. It is shown that the algorithms that rely on gradient information result in severe restrictions on the underlying problem structure and even on the hardware design for the sensor networks.

A suite of swarm intelligence algorithms are proposed for a variety of problems in decentralized sensor networks. In this first part of the dissertation, the algorithm is designed for a variety of sensor suites, model types, shifting the focus from the likelihood-ratio-test (top-down) based design to a more bottoms-up approach to solving the problem. In the second part, algorithm is designed to simultaneously optimize the two objectives, false alarm and miss, achieving a Pareto curve. In the third part, efficacy of the algorithm is demonstrated should the raw data from the sensors is available and the joint probability density function is hard to model. Again, the focus is a bottom-up approach to design, rather than the top-down design using a likelihood-ratio-test. Finally, the algorithm is adapted to a problem which has a different topology and a new algorithm is proposed that solves the coupled problem of threshold design and sequencing in tandem networks.

Access

Surface provides description only. Full text is available to ProQuest subscribers. Ask your Librarian for assistance.

http://libezproxy.syr.edu/login?url=http://proquest.umi.com/pqdweb?did=2071080741&sid=1&Fmt=2&clientId=3739&RQT=309&VName=PQD