Author

Ran Huang

Date of Award

2006

Degree Name

M.S. in Electrical Engineering

Department

Department of Electrical and Computer Engineering

Abstract

Swarm clustering is a biologically inspired technology, which mimics the behavior of an insect colony to solve unsupervised clustering problems. The unique mechanism of swarm clustering is self-organization which can form global patterns without centralized control. In this thesis, swarm-clustering models proposed by Deneubourg and by Lumer and Faieta are studied, implemented and improved by visualizing clustering results using true color representation. In nonlinear optical systems, self-organized patterns are often observed, such as hexagon formation in experiments with KNbO^Fe due to the photorefractive effect. Considering the particle feature of light, we propose to design a self-organizing system in which only a few simple rules are applied to local particle transported by the ant(s) from one position to another, and a hexagonal pattern emerges at the global level. The swarm-clustering idea may lead to an explanation of such self organized patterns without conducting complex optical experiment. In this thesis, we develop two different swarm-clustering schemes and successfully simulate the single and double hexagons formation, which are both observed in optical experiments. Parameter analyses are also performed to determine optimal conditions for self-organization.

Keywords

Self-organizing systems, Swarm intelligence, Cluster analysis

Rights Statement

Copyright 2006, author

Share

COinS