Some researchers at MIT have created a system that allows a pile of "smart pebbles" — tiny, cube-shaped robots — to create a 3D copy of whatever gets placed in it. Professor Daniela Rus and her student and co-author Kyle Gilpin will be presenting their results, complete with experiments using 10-millimeter cube prototypes, at the IEEE International Conference on Robotics and Automation in May.

In order to reproduce a 3D model, the first step is figuring out what you're reproducing, and the pebbles do this by figuring out if they're surrounded on all sides by other pebbles (in which case they're not touching the object to be copied), or if they're on a border somewhere (in which case they might be touching the object). Once the pebbles figure out where the model is they can send messages to neighboring pebbles to bond together and reproduce the same model. The actual bonding is done with electropermanent magnets on the cubes' sides that can be switched on and off, unlike conventional electromagnets, which require a constant current in order to maintain their magnetism. Finally, any unused pebbles remain in the heap, and shake free when the copy is removed.

The approach uses some ingenious algorithms that tell the pebbles what to do without a centralized command unit, a lot of computational overhead, or any of the pebbles having a complete map of the object. Rus and Gilpin believe that their research could be used for rapid prototyping, or producing replacement parts the way 3D printers are used today, although doing so would require the individual pebbles to be much smaller than their current 10-millimeter size.