Sanjoy Das presented a paper on his gene-spilling algorithm at the World Congress on Computational Intelligence in May 2002. He will co-organize and chair an international natural computing conference in North Carolina in 2003.
Engineers looks to nature to solve computer problems
Ants colonies an inspiration
By Mark Berry
Ordinarily, ants in a computer would be a bad thing.
But a Kansas State University professor brings organic life into the silicon world, using the inherent brilliance of nature to create better computer programming.
Scientists are constantly racking their brains to come up with new ways to design innovative programs. They can apply old concepts to new areas. They can use analytical approaches or intuition. But they cannot compete with evolution, said Sanjoy Das, assistant professor of electrical and computer engineering, who uses biological systems as a model for computer algorithms. An algorithm is a set of instructions used by the computer to accomplish its tasks.
"We have all these very complex problems, engineering and otherwise. Instead of trying to solve these problems by purely artificial means, I try to borrow things from nature and then try to incorporate them into problem solving," Das said.
An ant colony is one example. Each ant in the colony follows simple behaviors. Ants succeed in nature, however, because of the collective strategy they use and their ability to communicate with each other.
"The nest as a whole has some kind of intelligence. We can harness that intelligence and apply that to algorithms that can do optimization problems for you," Das said.
Das studies the biological system thoroughly, pulling out all its secrets and traits. In the ant algorithm, numbers represent each physical object in an ant colony. Those numbers are written into an algorithm, and voila an organic computation is born.
Das and Anil Pahwa, professor of electrical and computer engineering, are using the ant algorithm in electrical power distribution systems, like those that deliver power to your home or business. Das said the algorithm will improve the efficiency of the electrical systems, minimizing power loss. Das predicts that the ant algorithm will be several-fold faster than traditional algorithms in electrical power systems. They have also used ant algorithms to restore power more quickly after an outage.
Even bacteria, the simplest of organisms, have something to teach computer scientists. A bacterium passes on good genes by spilling its guts, so that bacteria nearby can absorb the improved genetic material. Das developed an algorithm modeled on the gene-sharing mechanism.
"Now we are going to apply it to some really complex problems, such as modeling electric motors and computer chip design," Das said.
He is also becoming interested in applying algorithms modeled after the behavior of birds flying in formation to engineering design.
Though Das earned his doctorate in electrical engineering, he has always been interested in biological systems. He did post-doctorate work on brain modeling and then learned about genetic algorithms.
"That opened up a window for me and I realized there are so many things out there that can be used in algorithms," Das said. "It was very exciting to me and when I look back, it makes a lot of sense. Why reinvent the wheel, when nature has evolved for such a long time? We have excellent problem solving strategies out there. Make use of them and apply them to the real world."
He and Don Gruenbacher, associate professor of electrical and computer engineering, are using the ant colony algorithm in the creation of adaptable computer hardware. Yes, in the future those little chips inside your computer may have the ability to reconfigure themselves. The two professors are working on "field programmable gate arrays" that adjust to different surroundings, much as an ant colony does. The chips can optimize themselves to the task they are given. Ordinary chips cannot.
"The microprocessor is set in stone. With the arrays, you can change the internal configuration," Gruenbacher said.
The inside of the array is like a large number of tiny switches that can be turned on or off. Programmers can change the configuration of the array by changing which switches are on and off. If the user wants to use the computer chip for ordinary computing purposes, it configures to a normal processor. It can then reconfigure itself to execute highly complex engineering tasks. Gruenbacher said NASA is interested in the arrays for long-distance missions. It would take years for satellites to reach distant planets, and if scientists wish to alter the mission in mid-flight, the chips can adapt.
The professors believe the hardware can run faster and more efficiently when the ant algorithm is implemented. The arrays can also be less expensive than traditional processors when mass-produced, Gruenbacher said.