By itself, this simple little puck-shaped robot is cute, but not revolutionary: It’s a few centimeters across, stands on three pin-like legs, moves a centimeter a second, and costs about $20. Put a thousand or so of these Kilobots together, and you have the largest robotic swarm the world has ever seen. 

Self-assembling robotic systems exist, but they’ve been limited to dozens, maybe a few hundred robots. Now, a trio of Harvard researchers led by Michael Rubenstein has programmed 1,024 Kilobots to organize themselves into various shapes, such as stars, a wrench, and letters of the alphabet. The work was published in Science this week. 

Kilobots were designed to mimic the behavior of a swarm of bees, colony of army ants, or flock of starlings. “Biological collectives involve enormous numbers of cooperating entities -- whether you think of cells or insects or animals -- that together accomplish a single task that is a magnitude beyond the scale of any individual,” Rubenstein says in an university release. 

These inexpensive robots have two little vibrating motors to help them slide across surfaces on their skinny rigid legs. Infrared lights let them talk to each other, but no farther than three bots away. 

The infrared transmitters are also used by the scientists to give commands to all the bots simultaneously. Once instructions are delivered, these autonomous Kilobots don’t need any micromanagement or human intervention to accomplish their task. First, four bots mark the origin of a coordinate system. All the others receive a 2D image to mimic, and they move into position using three rudimentary behaviors: follow the edge of a group, track a distance from the origin, and maintain a sense of relative location. 

The team started with an algorithm that relied on only those three collective behaviors of edge forming, gradient formation, and localization. They added more cooperative monitoring features when they noticed how some bots couldn’t work around collisions or broken down bots. "It's hard to make a perfect robot, and it's harder to make 1,000 perfect robots so any algorithm used to get a large swarm of robots to behave properly needs to account for faults in robots,” Rubenstein tells Popular Mechanics. With the new features, if a traffic jam forms or one moves off-course, nearby bots can detect this, and they’ll cooperate to fix it.

The robots don’t have access to a bird’s-eye view, and “there’s no centralized leader per se,” he tells Science. “Every robot is just talking to its neighbors and making its own decisions based on what it sees in its environment.” Here’s a series of images showing self-assembly of a "K" (for Kilobot) and a starfish: 

“The beauty of biological systems is that they are elegantly simple -- and yet, in large numbers, accomplish the seemingly impossible,” Harvard’s Radhika Nagpal says. With more advanced algorithms, these swarms could generate 3D shapes by attaching to each other and forming bridges or work together to respond to disasters. 

Here’s a video of the first thousand-bot flash mob:

Images: Michael Rubenstein, Harvard University