Stanford University biology professor Deborah Gordon loves ants. She’s got pictures of them on her walls, sculptures of them on her desk, and about 8 000 live ants running around her lab. She’s been studying them for more than 20 years. Recently, when she came to understand how colonies of harvester ants regulate foraging—that is, how they decide how fast to send out ant to gather food—she thought that the ant algorithm could apply to other types of networks.
Gordon went to Balaji Prabhakar, a professor of computer science and electrical engineering at Stanford University, with her idea. Prabhakar realized that the way her harvester ant network worked sounded a lot like Transmission Control Protocol (TCP), the algorithm that keeps the data traveling the Internet from bogging down when congestion increases or bandwidth decreases. The two set up a series of experiments to test the theory. They considered food in the ant world to be similar to bandwidth in the computer world; by manipulating the amount of food available they simulated changes in network bandwidth and watched to see if the ants adjusted in the same way a computer network operating under TCP would adjust, that is, by sending out less packets (or ants). The parallels were uncanny.
Had this ant research been performed in the early 1970s, when researchers were working on communications algorithms, it might have saved them a little time. Not that the researchers would have necessarily given the ants the credit. After all, quips Prabhakar, “the ants didn’t publish.”
Balaji Prabhakar: What the Internet and the “anternet” have in common is the algorithm itself that is used by senders of data.
Tekla Perry: Hi. Welcome to the Full Spectrum. I’m Tekla Perry, and I’m here with Deborah Gordon and Balaji Prabhakar and about 8000 ants. Long before researchers invented the Internet, ants invented something kind of like an anternet, and the two networks work pretty much the same way.
Deborah Gordon: I’ve been working for a long time on understanding how ant colonies regulate their behavior. And when I understood how harvester ants regulate foraging, I realized that the algorithm could be something that was used in other networking systems.
So an ant colony, like many other complex systems, works without any central control. There’s no one in charge, nobody tells the ants what to do, and they have to use the information they get from local interactions to decide what to do. And so I studied how ants use the rate and the pattern at which they meet to make decisions about what to do and how, in the aggregate, that allows the colony to adjust its behavior.
A colony regulates its foraging using the rate at which returning foragers meet outgoing foragers. An outgoing forager decides to go out depending on how quickly the ants are coming in with food. And because every ant that goes out searches until it finds food, the more food there is out there, the more quickly they find it and the more quickly they come back. So the rate at which foragers are returning is a measure of the availability of food.
Balaji Prabhakar: What the Internet and the anternet have in common is the algorithm itself that is used by senders of data in the Internet to detect the available bandwidth. It’s pretty much the same algorithm that the ants use to detect the amount of available food. The way the data network works is by speculatively sending some packets. And that’s analogous to the ants going out to forage. And then for every packet, the receiver of this information will then send an acknowledgement for every packet that they get. That is like the returning foragers. So the stream of packets going out and the acknowledgements coming back are sort of signals that [say,] “Okay, go ahead and send more.”
Deborah Gordon: So to really understand what’s going on, it’s important to be able to see inside the nest and to actually watch these interactions between the returning foragers and the outgoing foragers. This film was made with a videoscope. It’s about 5 centimeters inside the nest. And you see the ants coming up, and those are the outgoing foragers that are waiting for interactions with the returning foragers. So these are the outgoing foragers that are coming up into the tunnel just inside the nest entrance. And it’s the rate at which they meet ants coming in with seeds that’s going to determine whether they come out. So if you like, these are the data packets waiting to meet the acks [acknowledgments] to decide whether to go out.
To understand how colonies regulate foraging, we do experiments where we change the rate at which foragers return. We collect the returning ants as they’re coming back to the nest, put them in a box, and that brings down the rate at which foragers are returning. And also the rate at which foragers go out goes down because they respond to the returning foragers. And then when we let the returning foragers come back, when this red line goes back up, then the outgoing foragers start going out again. The ants are regulating their behavior according to food availability, but no ant knows how much food there is out there. So there’s no global assessment about the amount of food, and yet the colony can regulate very closely its foraging activity using just this rate of interaction.
Tekla Perry: Thank you both so much. I’ve been speaking with Deborah Gordon and Balaji Prabhakar at Stanford University about the anternet. For the Full Spectrum, I’m Tekla Perry.
Video: Tekla Perry; producer: Celia Gorman