How does one charge 1,000 robots? It would be a pain to plug them all in individually. Luckily, there’s an easier way with Kilobots. These little robots have round bodies about the diameter of a quarter, with a metal spring on top and three thin metal legs. To charge them, you push them — 10 at a time — against a long charging rack. They all charge at once, as long as each has its spring top and two of its legs touching the rack. “It’s kind of like a bumper car charging system,” says Mike Rubenstein, who uses a long stick to corral his Kilobots.
Rubenstein is an engineer at Harvard University and a member of the team that invented Kilobots (kilo meaning 1,000). The Harvard team began building simple, cheap robots more than three years ago, with the idea that the machines could be part of a cooperative swarm. They thought about issues such as how to charge a lot of them at once. Now, the engineers have improved the hardware and written software that allows them to send a single command out to 1,024 Kilobots; and in response, the little bots will start shuffling into place to form any solid shape researchers request:
It’s kinda creepy, until you realize it takes the robots 12 hours to make a shape. So they have their limitations.
That’s not to downplay Rubenstein and his team’s accomplishment. This is the first time anybody has been able to direct so many programmable robots at once. “They’re definitely on a point on the frontier where nobody else is,” Kevin Lynch, an engineer at Northwestern University who studies cooperative robots, tells Popular Science. Previous efforts used somewhere around 100 robots.
The Harvard work is part of a hot field in robotics. Basically, everyone in this field is trying to develop systems in which a large number of robots work together to accomplish a goal (“Robot swarm” is the popular term). The robots should be able to do so in response to just a few commands sent by one person. After that, the robots’ own basic algorithms need to be able to dictate how they should act individually to get the big job done. It’s maximum robot power with minimum human input.
Lynch offers an idea for what a cooperative robot system could do in the future. What if a person could send a single command to a fleet of drones to search a collapsed building for survivors? The drones could all enter at once, then use their algorithms to decide when to split up, to check different rooms, and to track which rooms have been checked already.
The challenge in building such systems is that there’s a limit to how complex each robot can be. For one thing, they must be cheap enough that someone could afford to own hundreds or thousands of them. The Kilobots cost about $14 in parts. Their only sensors are infrared ones that calculate the distance between themselves and their neighbors. That’s why the robots always march at the edge of the group, as you can see in the video. They’re keeping close to others because if they wander too far, they’re blind, unable to see their surroundings or calculate their positions.
Yet swarm robots must also be able to create sophisticated larger effects with their own simple capabilities. “It’s an interesting question of how do you go from a desired function — like if you want them to flock in a particular way or form a particular shape — how do you go from that to individual rules?” Rubenstein says. Coming up with those rules for the Kilobots, including rules for glitches such as a bot’s motor failing, was a major part of the research effort.
Although the Kilobots’ shapely 12-hour dances seem far from practical applications, Rubenstein says they could be a step toward robots that self-assemble into specialized shapes when needed. That kind of capability is especially of interest for doing projects in space. Engineers could send up parts of a larger satellite, for example, and count on the pieces to assemble themselves once they’ve been released into orbit. (Like Ikea furniture, but in space.) The smaller parts could be easier and cheaper to rocket into space than one large satellite.
Lynch adds that the Harvard work is basic research meant to push the field. “You can’t say well what it’s going to do for me tomorrow in a factory or how am I going to make money on this?” he says. “They’re asking new questions that are important.”
Rubenstein and his colleagues published their work yesterday in the journal Science.