clock menu more-arrow no yes

Filed under:

Watch a thousand robots execute the ultimate nerd choreography

New, 21 comments
Michael Rubenstein / Harvard University

There’s something unsettling about watching 1,000 robots execute a perfectly choreographed routine. The entire demonstration makes it a bit too easy to forget that humans, researchers at Harvard University, provided the algorithms that allowed them to accomplish various formations. And yet, these machines — tiny $20 robots that take five minutes each to assemble, for a total of 83 hours — are actually completely banal. In fact, according to the researchers, their capabilities are pretty abysmal.

"We created a swarm robot designed specifically to work in large groups, which was key to accomplishing this," said Michael Rubenstein, an electrical engineer at Harvard University who worked on the project, which is described in Science today, in an email to The Verge. "As a side effect, the robots are not very capable, and have lots of variability such as noisy distance sensing, and difficulties moving."

"the biggest group of decentralized robots working together to complete a complex global task."

Rubenstein is the lead researcher of a project aimed at creating a "thousand-robot swarm" — a collection of machines that he said is "probably the biggest group of decentralized robots working together to complete a complex global task."

Because the robots were so limited, however, the biggest challenge wasn’t the time necessary to build them, but coming up with an algorithm capable of governing the robot formations accurately. "We are partially motivated by swarms found in nature, including flocks of birds, groups of cells, and swarms of army ants — all working together to accomplish something bigger than any of the individuals can."

For now, the goal of the project seems to programming the swarm itself, rather than making it perform a task that might benefit humans. Eventually, the researchers may try to create larger swarms with smaller robots, which Rubenstein said will help them understand "how to control them even as the robots become less capable." And to add an additional layer of difficulty, he said, their next attempts will include robots that can attach to each other.