Accession Number:

ADA416407

Title:

An Artificial Immune System Strategy for Robust Chemical Spectra Classification via Distributed Heterogeneous Sensors

Descriptive Note:

Master's thesis

Corporate Author:

AIR FORCE INST OF TECH WRIGHT-PATTERSON AFB OH DEPT OF COMPUTER SCIENCE

Personal Author(s):

Report Date:

2003-03-01

Pagination or Media Count:

162.0

Abstract:

The timely detection and classification of chemical and biological agents in a wartime environment is a critical component of force protection in hostile areas. Moreover, the possibility of toxic agent use in heavily populated civilian areas has risen dramatically in recent months. this thesis effort proposes a strategy for identifying such agents vis distributed sensors in an Artificial Immune System AIS network. The system may be used to complement electronic nose E-nose research being conducted in part by the Air Force Research Laboratory Sensors Directorate. In addition, the proposed strategy may facilitate fulfillment of a recent mandate by the President of the United States to the Office of Homeland Defense for the provision of a system that protects civilian populations from chemical and biological agents. The proposed system is composed of networked sensors and nodes, communicating via wireless or wired connections. Measurements are continually taken via dispersed, redundant, and heterogeneous sensors strategically placed in high threat areas. These sensors continually measure and classify air or liquid samples, alerting personnel when toxic agents are detected. Detection is based upon the Biological Immune System BIS model of antigens and antibodies, and alerts are generated when a measured sample is determined to be a valid toxic agent antigen. Agent signatures antibodies are continually distributed throughout the system to adapt to changes in the environment or to new antigens. Antibody features are determined via data mining techniques in order to improve system performance and classification capabilities. Genetic algorithms GAs are critical part of the process, namely in antibody generation and feature subset selection calculations. Demonstrated results validate the utility of the proposed distributed AIS model for robust chemical spectra recognition.

Subject Categories:

  • Chemical, Biological and Radiological Warfare

Distribution Statement:

APPROVED FOR PUBLIC RELEASE