Researchers at MIT have developed small building blocks called smart sand which can duplicate any shape they get in contact with.

Smart sand is MIT’s realization of a programmable matter system. Programmable matter is a concept which originated in the early 1990’s referring to a substance that can change its physical properties by means of processing information (Wikipedia). Within 2 decades science fiction has already become science fact.

Professor Daniela Rus director Distributed Robotics Laboratory at MIT’s Computer Science and Artificial Intelligence Laboratory and her student Kyle Gilpin will present a paper with the latest developments at the IEEE International Conference on Robotics and Automation in May.

Their ultimate goal is to develop a universal smart sand toolkit which can replicate almost any object autonomously. The toolkit consists of nothing other than a bag of smart sand. When an object is dropped into the bag the building blocks form a duplicate in seconds. When the duplicate is no longer needed, it can be chucked back into the bag where it disintegrates back into blocks and can be re-used to form something else.

So how does it work?

There are three key elements to the smart sand: electro-permanent (EP) magnets, algorithms and communication. Each ‘sand grain’ or module is a 12mm cube with EP magnets attached to four of its sides. The magnets are strong enough to attract other modules from a distance and once connected they can transfer data and power between the modules.

The EP magnets can maintain their magnetic field with zero power input. They only need an electric current to change their state into on (attraction) or off (no force exertion). So once the cubes are latched together they do not need a constant input of electricity to keep in shape.

The novelty of this programmable matter system lies in its algorithms. Generally researchers who work on program-driven self-assembly make the modules form shapes in a single pass. That is to say, the assembly starts with a single module attracting other modules which are only allowed to attach themselves at specific locations. The structure self-assembles block by block.

Smart sand, on the contrary, starts out as a single substance consisting of all the cubes latched together, forming a close-packed crystalline lattice of modules. Each cube is attached to eight neighbors except those on the surface which miss one or two. The information about missing neighbors is communicated amongst the cubes to determine collectively where the surface is.

When an object is inserted in the substance those cubes surrounding it also start reporting missing neighbors. This information is shared to establish the parameter of the embedded object. The parameter is than replicated elsewhere in the lattice. When that’s done the cubes not forming the replica switch off and detach from the replica. This process is visualized in the MIT News Office video embedded below.

Smart sand is not primarily a self-assembling technique but a self-disassembling one. The reason for this different approach is scale. At 12mm the smart sand grains are rather smart pebbles but Professor Rus and her team are aiming for grain-sized in the future. That means they need to develop efficient algorithms that do not require a lot of computing power.

The single pass self-assembly technique requires that every cube has digital models stored  in its memory of all objects it needs to be able to replicate.

That’s why smart sand can duplicate an object but can’t produce one that is not present. It can, however, enlarge the replica. Allowing for a miniature model to be put in the universal smart sand toolkit and getting a full-sized replica out.



For more technical details see: groups.csail.mit.edu

Photo: M. Scott Brauer from web.mit.edu/newsoffice