Date of Award

3-19-2009

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Department of Electrical and Computer Engineering

First Advisor

Gilbert L. Peterson, PhD

Abstract

Military and industry are moving toward every device being network enabled and connected for reliable availability of communication and information. To make this type of system a reality, the devices must be capable of forming a network topology on their own in a dynamic environment to ensure that the correct information reaches a desired location and on-time. This research presents three contributions for solving highly dynamic (i.e. drastic change within the network) Multi-commodity Capacitated Network Design Problems (MCNDPs) resulting in a distributed multi-agent network design algorithm. The first contribution incorporates an Ant Colony Optimization (ACO) algorithm Ant Colony System (ACS) to solve the static MCNDP with weak constraints. Second, a new algorithm is developed and has the capability to dynamically adjust its exploration parameter of the solution space. This enhanced algorithm converges quickly and automatically adjusts to the dynamically changing network environment. Third, a distributed approach is created replacing the previous centralized solver. The distributed algorithm produces comparable results, but more importantly calculates the network topology in less than 20 percent of the computation time.

AFIT Designator

AFIT-DCS-ENG-09-06

DTIC Accession Number

ADA495713

Download

Share

COinS