DURHAM, N.C. -- Using $6 million in U.S. Army research funding to be channeled
through Duke University, investigators at five different institutions will
evaluate new electronic surveillance measures against a hidden and deadly
menace: the concealed land mines currently endangering the populations of
more than 60 nations.
Researchers at Duke's School of Engineering will join others at Caltech, Georgia Tech, Ohio State University and Stanford University to explore innovations in mine detection ranging from a microelectronic chemical-sniffing "nose" and through-the-air ultrasound to ground-shaking seismic waves and unique mathematical and computational aids.
The collaborators plan 12 different research projects to better detect buried -- and often diabolically clever -- military booby traps that often kill or maim innocent noncombatants.
The projects will fall under three research areas:
- Chemical-sensing processes that mimic smell;
- Radar, magnetic, infrared and sound sensors;
- Sensor-information processing.
"For example, the mines Iraq placed had electrical properties which were almost exactly the same as the sand," Carin said in an interview. "That made them extremely difficult to detect using electromagnetic means. And in Bosnia many of the mines are made out of wood; you just can't detect them with a metal detector."
According to a Department of Defense research report, estimates of the number of uncleared mines worldwide range from 85 million (by the U.S. Department of State) to 105 million (by the United Nations) spread over 62 nations. Between 500,000 and 1 million new mines are believed to be deployed each year. And somewhere between 10,000 and 100,000 people are said to be killed or maimed by them annually.
Before the new MURI even begins, enough recent research has been done on mine detection to fill a 1,036-page book on Carin's desk with summaries of work on various kinds of radar, X-ray, magnetic, infrared, ultraviolet and laser devices.
It would require an estimated $56 billion to clear out all the world's unexploded mines, said Richard Fair, a Duke professor of electrical and computer engineering who is also involved in the initiative.
Fair's expertise is in the fabrication of semiconductor microchips as well as tiny machines known as microelectromechanical systems (MEMS). His team will collaborate with Caltech chemists developing advanced electronic odor-sniffing polymer strips -- branded a "nose on a chip" -- that can identify buried explosives much like specially-trained dogs.
The Caltech team, led by chemistry professor Nathan Lewis, has already developed films sensitive enough to tell "beer from wine from hard liquor, dead fish from live fish, and rose oil from garlic" -- as well as being hypersensitive to TNT-related explosives, according to a research outline.
These special "olfactory" sensors discriminate between those odors by measuring how different chemical vapors change the polymers' electrical resistance. "It's not really on a chip yet, but it will be," Fair said.
Fair's team will help Caltech's nose "sniff" better by researching new ways to find and deliver to the sensors tell-tale vapors from the buried mines. The researchers will capitalize on the Duke School of Engineering's expertise in ultrasound -- a technology that uses high-pitched sound waves to create images of hidden internal anatomy.
While ultrasound waves used in medicine or industry are typically transmitted through fluids or bodily tissues, Fair's group has been working on a tiny MEMS device that can send and receive ultrasound in the open air. A strong enough ultrasound signal sent that way could penetrate the ground and detect otherwise unobtainable signatures of buried mines there, he said.
"We would have a steering-and-location MEMS ultrasound system," Fair added. He proposes building a mobile device that would guide odor-detecting microchips to locations that ultrasound reflections suggest harbor mines. Once on location, the ultrasound beams would also be intense enough to stir up underground chemical particles for identification as explosives.
Then another MEMS detector that Fair proposes building would suck up those chemical particles, weigh them, and analyze them "so that the nose on a chip would have additional sensitivity," he said. Fair's calculations show a minuscule balancing beam microengineered out of silicon could easily weigh particles as light as 10 trillionths of a gram (one gram weighs .035 ounce).
MEMS manufacturing techniques are similar to those used to fashion microchips. That means MEMS mine-detection devices could be fabricated at MCNC, a microelectronics and computer research center in Research Triangle Park founded by area universities and the state of North Carolina, said Fair, a former MCNC vice president.
Carin, a radar expert who began researching radar's applications to mine detection during his six years at Brooklyn Polytechnic Institute, said his new Ohio State collaborators "are some of the real pioneers in this. We saw this research initiative as an avenue where we can finally work together."
The Ohio State team, led by senior research scientist in electrical engineering Eric Walton, will use an experimental test range to evaluate advanced strategies for using ground-penetrating radar technology in mine detection.
Meanwhile, Georgia Tech investigators, led by associate engineering professor Waymond Scott, will add a new seismic approach to radar detection, Carin said. "They're going to shake the ground to try and produce a seismic, or acoustic wave -- a very tiny earthquake, if you will."
Using either a loudspeaker or a ground-contact vibrator, they will look for differences in the way buried mines and surrounding soils respond to the shaking. The Georgia Tech researchers will then see if those differences can be highlighted by radar beams aimed at the shaking surface soils. The hope is for improvement on using radar alone, which has difficulty penetrating much below the ground surface, Carin said.
OSU researchers also will join Duke electrical and computer engineering professor Erol Gelenbe and assistant professor Leslie Collins in exploring the potentials of infrared radiation for detecting mines.
Invisible infrared radiation, which humans can sense as heat, is the basis for special night vision goggles the military uses to detect after-dark troop movements. According to a research initiative proposal, the most sensitive infrared detectors might also register subtle surface temperature effects produced by buried mines.
In joint work with Georgia Tech, Collins also will work on strategies to improve synthetic aperture radar, which uses computer technology, fog-penetrating microwave frequencies and a moving antenna to build up radar images that grow sharper over a period of time.
Another mine-detection approach -- spraying the soil with frigid liquid nitrogen or "highly cooled" air -- will be explored by Carin with Caltech and Ohio State researchers. That research will evaluate whether freezing the soil would increase radar penetration or cause other changes detectable by infrared or odor-sniffing devices.
Carin and Duke electrical engineering professor William Joines also will help develop a new magnetic-detection technology -- a pulsed magnetic sensor. "Magnetic sensors right now are basically the state of the art in the Army right now," Carin said. "And, working with people from industry, we can actually do much better than is currently being done."
They will assist EG&G of Albuquerque, N.M., one of three "industrial partners" in the MURI and the prime developers of the new magnetic sensor. EG&G and the other two firms, Hughes Aircraft Co. of Malibu, Calif., and Northrop Grumman Corp. of Baltimore, are providing technical support for a number of the initiative's projects but will receive only minimal government funding.
Stanford's contribution to the MURI will be mathematics professor George Papanicolaou, who will provide his expertise to several different projects.
The initiative's third research area -- "sensor fusion and signal processing" -- will team up Gelenbe, the chairman of Duke's electrical engineering department, with Carin, Collins and other Duke faculty, as well as investigators at Ohio State.
Sensor fusion involves merging information from various sensors in an "intelligible" way. "You have these different pieces of information coming in," Gelenbe said. "You have to combine that sensing information and then make intelligent decisions."
The researchers also will develop methods to decide the safest and most effective locations to place the humans or robots deploying the high tech sensors in the field.
"You can develop some of these decisions ahead of time," Gelenbe added. "But some strategies will have to be designed for 'real time' implementation, in the sense that the commander of a demining unit will have to make decisions shortly before -- or even during-- deployment."
Gelenbe will work closely with Duke computer science professor John Reif, and with experimental psychology professor John Staddon and associate professor Nestor Schmajuk, to study optimal ways to deploy sensors and other mine detectors in the field.
To sift through the sensor data, Gelenbe will use neural networks, computer systems that emulate how networks of nerve cells process information. More specifically, he will employ a "recurrent random neural network" model developed at Duke to tackle difficult problems in data interpretation.
He also will use a novel genetic algorithm of his own invention that can generate more precise answers quickly. Genetic algorithms are special mathematical tools that "evolve" the best solutions to problems by weeding out "unfit" ones.
Research to develop signal-processing systems for mine detection will involve using computers and mathematical models to determine whether a sensor's often-abstract input really does reveal a land mine's presence.
"We're also modeling the uncertainty, and I think that's something that is very novel," said Collins. "If you can somehow characterize the uncertainty about the model, then you can actually do a much better job."
Carin said "the real challenge is to take all that data and try to make a decision. It's analogous to how humans use our hearing, smell, taste and touch. We don't just use one sense."
The researchers will also develop methods to decide when not to use a given technology. "Radar has been used for decades to try to find mines," Carin added. "It actually works very well in some scenarios, and very poorly in others. The real challenge is to make sure you use radar when it's appropriate. And then when it's inappropriate don't even try to use it. Because you're just asking for trouble."