Dr. James McLurkin has a swarm of robots. Individually, they're not that smart, but throw a crateful of them on the floor, and they'll behave in some very complex ways, like the bumblebees that inspired them. Gizmodo got to see them in action, and while the wee machines are undeniably adorable, they represent some serious future bot capabilities.
The quiet before the swarm: robots waiting to be powered-up.
Dr. McLurkin, a professor of computer science, runs the Multi-Robot Systems Lab at Rice University. He and his team research distributed algorithms for multi-robot systems. In other words, using the combined abilities of several rather simple robots to perform complex tasks. Dr. McLurkin has spent the past three years developing Robot Swarm, an exhibit of his hive-mind bots set to debut at Manhattan's Museum of Mathematics in early 2015. This week, Dr. McLurkin gave a sneak preview of the exhibit, and Gizmodo was there.
"To a computer scientists, a bee's behavior looks like a flowchart," Dr. McLurkin said at the demonstration. "Nature is full of complex group behaviours carried out by simple individuals." He gave the example of a group of ants picking up a potato chip from a picnic site: Though they might face different directions, the ants all agree which way to carry their freight.
Dr. McLurkin's bots are the size of a small flower pot. Each one wears an infrared LED and four IR sensors, one on each corner, telling each bot where its nearest neighbour is. Individually, each bot only "knows" where it is in relation to its nearest colleague, but give six of those critters an algorithm to work within, and you'll see some complex, coordinated bot behaviour.
Dr. McLurkin used a modified PlayStation controller to toggle through different behaviours for his swarm of robots. In one mode, each bot was told to draw a path to its nearest neighbour, shown above. The custom-built, light-sensing LED floor visualized these paths for the audience, using data beamed from each bot's undercarriage. Another task had each bot figure out how it would hop from neighbour to neighbour to reach the "leader" bot, shown below.
Then Dr. McLurkin got the bots moving. In one task, the bots were told to keep the leader bot, controlled by Dr. McLurkin, over their left shoulder at all times. After a few seconds of scurrying confusion, the bots began orbiting around the leader like planets around a sun. In another, they were told to line up by assigned number, lowest to highest.
The robots did all of this knowing essentially nothing about the world around them—only the location of the nearest kin. Dr. McLurkin sees a future where bots like this use their not-quite-intelligence to do dirty or dangerous tasks. "What if we sent 20 robots to look for hot spots in a forest fire? Or 200 robots to look for earthquake survivors? Or my favorite robot job, sending robots to Mars. We've got two there now, but what if it wasn't two, but 2,000?"
The exhibit catered to kids without talking down to them. The minibots squeaked, honked, and whistled like tiny R2D2s, singing "Hi Ho, Hi Ho" and "The Hokey Pokey" during different tasks.
The kids were fascinated by these toy-like bots, but not just because they looked like little cartoon characters bopping around on a disco floor. They wanted to understand why they worked the way they did, a task Dr. McLurkin was happy to explain.
Summer Swarm was a one-time sneak peek—the full Robot Swarm exhibit will open at the Museum of Mathematics in December. We can't wait to see what it looks like.
Images and GIFs by Nicholas Stango