DAVID POGUE: What happens when you move beyond having just one robot and, instead, have a swarm? In the future, swarms of robots operating as a team might build our skyscrapers or map uncharted areas or scout out victims in disasters, as robotic search and rescue teams.

But in order to do any of that, engineers must solve a problem nature solved eons ago: "How do you get a group of individuals to work together as one?"

In nature, swarms often behave as if they have a collective intelligence. Whether it is fish schooling in the sea, or birds flying in a flock, the members act in unison, without anyone, apparently, in charge.

Some of the achievements built out of this swarm intelligence are awe-inspiring, like this murmuration by thousands of starlings, or these complicated towers built several feet high by blind termites.

So what can we learn from behavior in nature about creating robotic swarms?

Vijay Kumar and his students at University of Pennsylvania have been wrestling with the problem. They use a fleet of hand-sized quadrotor robots, which they've learned to manipulate with impressive control. They can play the theme from James Bond or put on a lightshow. In both performances, the quadrotors are individually controlled by a central computer.

But they've also built some computing power into individual robots, so they can think for themselves, like figuring out how and when to fly through a tossed hoop.

Now, Vijay is taking the next step, developing software that will allow the bots to work together as a swarm, a team that can do more than any single flyer can.

One flying bot, pretty cool; eight flying bots? It gets a little "swarm" in here.

VIJAY KUMAR (University of Pennsylvania): So what you see here is these robots are commanded to rise into a swarm. They're asked to form patterns, three-dimensional patterns, and then the robots figure out what point in the pattern to step into and how to coordinate with their neighbors.

DAVID POGUE: Oh, so the master computer doesn't say, "You be in the corner." It's just saying, "Be a rectangle," "Be a circle," but they have to decide how to execute that?

VIJAY KUMAR:Right.

DAVID POGUE: Well, a central computer could control each of the eight robots individually, telling them where to go, but Vijay wants a system that scales up, and with more robots, no computer could keep up.

So, instead, he's taken inspiration from swarms in nature, and developed three guiding principles.

First, as much as possible, just as in nature, each robot thinks for itself. Second, each robot acts primarily on local information it gathers, the way a bird in a flying flock probably pays attention only to its immediate neighbors to know where to go. Finally, no one robot is in charge. They're all interchangeable, so that if one breaks down, the group continues.

To test out those principles, Vijay turns his fleet over to me and lets me experiment.

The flyers know they're supposed to make a circle. As I add them, one at time, you can see it take shape. Or, I can randomly pluck one out of the air—proving none is essential—and put it back somewhere else. Its neighbors adapt.

Of all the possible applications, Vijay sees a big future for swarms in search and rescue. And he shows me how it would work.

VIJAY KUMAR:So, imagine you have a victim, you can imagine robots wandering around looking for, maybe, a cell phone signal that might tell you where the victim is.

DAVID POGUE: Find that, robot drones!

BENJAMIN CHARROW (University of Pennsylvania): I'm telling these bots to move.

DAVID POGUE: The numbers of the bots?

BENJAMIN CHARROW: The numbers of the bots. They're all moving around.

DAVID POGUE: All right, there they go. They've chosen independent routes.

In this demonstration, run by one of Vijay's grad students, the robots roam the floor, measuring the strength of a signal transmitted by our lost victim's cellphone. They share their readings, creating a map.

In effect, the transmitter is saying, "Warmer, warmer."

BENJAMIN CHARROW: Cooler, cooler. But the robots know not to trust any one sensor reading, too much.

DAVID POGUE: The swarm of robots cut the overall search time, by gathering information faster than a single robot could, leading to the rescue of our lost little guy.

VIJAY KUMAR:Well, right now it's a lab study. But this sort of illustrates one of the directions in which we want to go, which is you take very simple robots with very simple sensors, so they're inexpensive, you put them together, and suddenly you have the benefits of these robots collaborating to do things that they individually cannot do. And that illustrates what swarm technology can really do for you.