Annual Report 2007

Improving Missile Warning Systems with Genetic Algorithms GTRI is helping improve missile warning systems. An Air Force technician, left, loads a chaff dispenser while an AC-130H Gunship, above, jettisons flares to distract missiles. G threats. TRI researchers are using genetic algorithms to improve the missile warning systems that alert pilots to approaching Senior research engineer Greg Rohling is optimizing the detection software that distinguishes missiles from background signals. Many constants in the code are good candidates for optimization, according to Rohling, who aims to detect all threats not identified by existing software – while not increasing the number of false alarms. “The code might say ‘if x is greater than three, then do something.’ That three is a magic constant,” said Rohling. “Would changing the constant to 3.1 or 2.5 allow the program to detect more missiles?” With funding from Warner Robins Air Logistics Center and Naval Air Systems Command (NAVAIR) Tactical Aircraft Protection Systems (PMA272), Rohling found hundreds of these constants in the software. “Think of each constant as a knob. I can turn the hundreds of knobs a little bit and every time I might get a different detection rate,” said Rohling. “But evaluating all of the knob combinations would take years, and we don’t have that kind of time.” To evaluate the best value for each constant, Rohling uses a computing concept called genetic algorithms to speed up the optimization process. By mimicking the same processes Mother Nature uses, such as selection, mating and mutation, genetic algorithms evolve solutions to problems to find the best one. He begins with many different sets of knob combinations; each set is called an individual. Principal research scientist Ed Patterson creates missile data that allows each individual to be tested using computer simulations. After a round of simulations, poor performing combinations of constants are thrown out and remaining individuals can be selected, mated or mutated. Individuals that performed well are selected to move on to the next round. Then, based on their performance, the remaining individuals have a certain probability of mating and a new individual is created by selecting half of the constants from one individual and half of the constants from another individual. Mutation involves tweaking a few knobs in one individual to create a new individual. New individuals are tested until the detection rate no longer improves. The team tested more than one million different constant combinations to find the best. “This optimization provides a quantitative way of measuring the success or failure of the computer program that detects missiles,” explained Rohling. Principal research engineer Jeff Hallman and research engineer Garth Girman verified that the optimized software remained compatible with the system hardware by testing thousands of recorded test range files and synthetic missile shots. “We found that the optimized program improved the hardware’s missile detection rate and did not increase the false alarm rate compared to the current software,” said Hallman. The optimized software will be submitted to Alliant Techsystems, which subcontracted GTRI for the testing part of the project. Later, it will be tested by the U.S. Navy. The GTRI researchers working on this project are among more than 50 teams performing research and development activities through SENSIAC, the Military Sensing Information Analysis Center. SENSIAC is operated by the Georgia Institute of Technology and provides information and research services to all defense-related elements of the U.S. government. www.gtri.gatech.edu 1 http://www.gtri.gatech.edu

Table of Contents Feed for the Digital Edition of Annual Report 2007

Annual Report 2007 Contents From the Director Research at GTRI Defense and Security Health and Human Systems Energy and Environment Manufacturing Technologies Information and Communication Technologies GTRI Around the World Interdisciplinary Research Centers The Laboratories of GTRI Georgia Tech Ireland GTRI by the Numbers GTRI and Georgia Tech Supporting Breakthrough Research GTRI Professional Education The People of GTRI

For optimal viewing of this digital publication, please enable JavaScript and then refresh the page. If you would like to try to load the digital publication without using Flash Player detection, please click here.