These are Kilobots. They're fairly simple little robots about the size of a quarter that can move around on vibrating legs, blink their lights, and communicate with each other. On an individual basis, this isn't particularly impressive, but Kilobots aren't designed to be used on an individual basis. Costing a mere $14 each and buildable in about five minutes, you don't just get yourself one single Kilobot. Or ten. Or a hundred. They're designed to swarm in the thousands, although the Harvard group that's working on them is starting out with a modest 25:
We've seen lots of examples of swarm robotics, but what we decide to call a "swarm" often isn't, really. There is (or should be, at any rate) a distinction between a group of robots cooperating on a task and a true swarm of robots, and for the purposes of this article, I'm going to arbitrarily assert that a group of robots turns into a swarm of robots when you can't easily count how many individual robots there are. So like, these swarming MAVs? Not really a swarm. Swarmanoid? Not a swarm yet. Swarm bots are getting closer. What definitely makes the cut are projects like RoboSwarm and FlyFire, which use anywhere from hundreds to thousands of small robots all at once.
There's a lot you can do with gigantic swarms of robots, but there are two big obstacles to deploying them: programming, and charging. If you can't figure out a way to do these things efficiently (i.e. not on an individual basis for each robot), it negates a big part of the swarm appeal. In the case of the Kilobots, they can all be programmed at once with an infrared controller, and to charge them, the bots can simply be sandwiched between two conductive surfaces. The fundamental idea here is that any interaction with a robot swarm has to be scalable, such that an increase in the number of robots in the swarm doesn't result in an increase in the amount of time it takes to interact with the swarm.
I should point out that the other big obstacle to robot swarm deployment is price, which is why kilobots are deliberately so cheap: at $14 each, a thousand robots is actually an achievable number with a modest grant, which is something that probably has not been possible before. Generally people who want to experiment with large swarms have had to be content with computer simulations, which is fine, but at some point you have to try things out in the real world (or as close as you can get in a lab), and Kilobots can make that happen.
The Self Organizing Systems Research Group at Harvard is planning to expand their Kilobot collective to 1024 robots, and then they'll teach the swarm to demonstrate behaviors like self-healing and collective transport. Better hide your kids. Also, for the record, I'm pretty sure it's "Kilobots" and not "kill-o-bots." But who really knows until it's too late, right?