A sophisticated robot is a smarter robot, but it’s also much more expensive and harder to build. A simple robot is a dumber robot, but get many of them to work together and you can accomplish something greater than the whole. Publishing today in Science, Harvard scientists have found that balance between simple and sophisticated, creating the largest swarm of self-organizing robots to date.
The future of robotics probably won’t be large, humanoid automatons building our cities and exploring dangerous environments. Rather, researchers are looking to the intelligence of the swarm. Imagine if we could one day program legions of microscopic robots to act as a kind of machine matter. They could run like the cells of a super-organism, making tools or objects at a whim, accomplishing tasks, and re-shaping themselves like a 3D printer without the printer.
We aren’t quite at programmable robot goop yet, partly because it’s hard to get a lot of robots to work together. Currently, most efforts to harness a robot swarm fail to get 100 robots or more acting as one. This new study out of Harvard boasts 1,024.
The robots themselves — dubbed Kilobots — are of the simple and dumb variety, but they get the job done. Equipped with three vibrating legs (they move around like you cell phone does on vibrate), a processor, and infrared transmitter/receiver, these penny-sized robots communicate with each other via infrared light. What makes them special though is the algorithms that govern them. Each of the 1,024 robots only costs about 14 dollars each, so the secret behind their being able to form any two-dimensional shape is indeed the infrared language they blink to each other.
To form a shape, four “seed robots” are placed at the point where an imaginary origin of a X and Y coordinate system would be. Then, one by one, the Kilobots buzz past each other, measuring relative distances like how far they are from the next robot or from the edge of the shape they are supposed to make.
Once a Kilobot is inside the shape the group intends to make, it follows the outside edge of the initial mass of robots until it bumps into the Kilobot in front of it or it reaches the coordinates of the edge (based on those four seed robots). All the swarm needs are the first set of instructions from the researchers, and off they go without any further assistance or programming.
The next step is better programming and smaller, more adaptable robots. Once the algorithms get faster and the robots can communicate more efficiently, we will be able to direct more and more of them. That’s when we start thinking about how to get them to link together to form tools, structures, or Voltron (hopefully).
—
IMAGE: Michael Rubenstein, Harvard University
This is very cool, but I don’t understand how the robot swarm know what the desired shape is?
The edge coordinates for the desired shape are programmed in at the beginning. So they start with the target shape (as coordinates) and a starting point (the seed robots), and from there the algorithm is able to complete the process autonomously.