THB, Banditos, Wayward and more confirmed for Cosmic Cocktail!

Electronic ants search the Internet


DALLAS -- The world could be a very efficient place if more people would just stick their heads in the sand -- and take a look at all the ants.

To you, perhaps, ants mostly mean spoiled picnics. But to a handful of scientists who want to improve airplane travel, telecommunications networks and Internet traffic, ants are pure inspiration.

The idea is simple. Ants do some pretty intricate tasks. Even without the benefit of training seminars or personal organizers, thousands of ants can figure out how to cooperate to build colonies, find food, tend to their young and remove their dead.

Scientists think if they can learn how the ants do so much and adapt so well, they might be able to apply the same insights to the world of humans. And since ants have had millions of years to perfect their skills, imitating them is a risk worth taking, scientists say.

"They are a very successful family, and it's worth paying attention to them," says Craig Tovey, a computer scientist and biologist at the Georgia Institute of Technology in Atlanta. "They certainly are coping with an uncertain world."

Developing artificial devices based on nature is not a new idea for scientists, says Eric Bonabeau, a computer scientist at the Santa Fe Institute in New Mexico who studies ants. Scientists working on radar have studied echolocation in dolphins. And the brain inspired researchers to create computer programs called neural networks.

But the idea of studying ants to help streamline the lives of people has been around for less than a decade. Only a few dozen scientists around the world, most of them in Europe, are working in the field, known as ant-colony optimization. The first conference ever on the topic was held just last month in Brussels, Belgium.

And while scientists outside the small field aren't familiar enough with it to predict how it will pan out, they find the idea intriguing.

"The promise that I really see in what these guys are doing is that they're hopefully discovering some general principles that the ants appear to be following," Tovey says. "And that these principles are ways of dealing with complex problems and with a lot of uncertainty."

One of the ants' most valuable skills -- at least to the computer scientists -- is the ability to find the quickest route between two points. Even when an obstacle -- such as a fallen stick or a puddle -- suddenly appears, ants are undaunted and quickly find the next-best path.

These ant-sized problems have parallels in the world of humans. When telephone calls travel from coast to coast, they may be routed through several nodes on the network. But if telephone traffic ties up a particular node, sometimes it's necessary to reroute calls. The same goes for Internet traffic, or even airplanes affected by bad weather in a particular city.

The ants' solution to coping with traffic jams is to plan ahead, Bonabeau said. When ants go marching -- carrying food, or cleaning out their nest, for instance -- they deposit a trail of chemicals called pheromones. Other ants can follow that trail. But while most of the ants are following the main pheromone route, a few ants stray from the path and explore other options, also depositing pheromones as they go. So if traffic jams, the ants are prepared.

"They maintain a set of alternate routes all the time," Bonabeau says. "This exploring sets up a backup plan."

That kind of adaptability is what's needed in routing traffic through the unpredictable Internet, says Marco Dorigo, a computer scientist at the Free University of Brussels. For instance, he says, it's impossible to predict what Internet traffic will be between Boston and London on any given day, at any given hour.

"Basically, nobody knows how to do that," he says.

But if there were ants -- virtual ants, not real ones -- that continually explored the Internet, Dorigo says, backup routes could be monitored.

Dorigo, along with Gianni Di Caro of the Free University of Brussels, has devised such a system, which he calls "AntNet." The paper describing AntNet, to appear in the Journal of Artificial Intelligence Research, compares AntNet with other Internet routing programs invented by other scientists.

In AntNet, the virtual ants are little snippets of computer code that zoom around the network. Instead of dropping chemical pheromones, the virtual ants, also called agents, leave behind bits of electronic information to enlighten other ants that may be passing through later.

As they move from node to node, Dorigo says, the virtual ants bring back traffic updates and can also deposit electronic information that actually changes the direction of Internet traffic.

And like real ants, the virtual ants work as a colony.

"If you have just one, the system doesn't work," Dorigo says.

AntNet addresses only the routing of information in networks, Dorigo says, and not other necessities such as error detection and the ability to retransmit lost information.

But AntNet did well when compared in simulations with other routing programs, says Craig Boutilier, a computer scientist at the University of British Columbia in Vancouver. The scientists gauged how much information AntNet could handle and how long it took to get to the destination. However, Boutilier says, it's hard to tell whether the system would work on the real Internet.

"Their claims are pretty compelling," he says. "But whenever you set up an experimental model, you have to make simplifications, so who knows how reasonable they are."

And before ants would be let loose on the Internet, Boutilier says, the Net would have to be ready for the ants.

"If a node on the network doesn't provide the means for an ant to leave their little electronic bread crumbs, then the whole thing breaks down," he says. "You really need the entire network to allow the ants to do their thing."

The ant approach to solving human problems is only in its infancy, some scientists note, too soon to know whether it will prove superior in the long run.

And even its advocates predict that it might take a while to catch on.

"I'm not sure that this algorithm will be used in networks that soon, and the main reason is people's mentality," Bonabeau says.

"Managers would rather live with a problem they can't solve than work with technology they don't trust. Being replaced by ant-like agents is scary for lots of people."

Pub Date: 12/05/98

Copyright © 2019, The Baltimore Sun, a Baltimore Sun Media Group publication | Place an Ad