MIT imagines self-forming "smart sand"

A team at MIT has come up with a vision of what it calls self sculpting sand – “smart sand” – theorising that using a new algorithm, by placing a model in a box of sand, the sand could one day assemble itself into a large scale replica. 

A research project at the Distributed Robotics Laboratory, at MIT’s Computer Science and Aritificial Laboratory, is announcing a paper including algorithms which could eventually enable such a technology. This smart sand would be using a “subtractive method”, similar to chipping away at a stone carving, rather than an additive method. 

MIT says that the equivalent example would be the starting steps of a sculptor – turning a block of stone into a final product. The researchers theorise that individual grains could talk to each other to make themselves into a 3D object, while unneccessary grains would fall away. When that 3D object had served its purpose, it could then be returned to the sand heap – with the grains then detaching from one another, ready to make a new shape altogether.

According to MIT, the biggest challenge so far in developing such technologies is that individual grains of sand do not have enough computational resources. However, co-author of the paper Daniela Rus said that if each grain was able to store a digital map of the final object, it takes some challenges away from coming up with the algorithm. “We would like to solve the problem without that requirement, because that is simply unrealistic when you’re talking about modules at this scale – we’d like to not have to know ahead of time what our block looks like,” Rus said. 

To imagine the algorithm, MIT suggests picturing each grain as a square in a 2D grid. If some of those squares are missing, such as in the shape of a footstall, this is where the physical model is embedded. The grains can pass messages to each other to figure out where there are missing neighbours. Missing neighbours will either be in the perimeter of the sand heap, or the perimeter of the embedded shape. Then, the grains which surround the embedded shape are able to identify themselves, passing messages to other grains, which can identify themselves as the perimiter of a duplicate object.

So, if the duplicate is supposed to be a large version of the shape, the square surrounding the embedded shape will map themselves to 10 squares of the duplicate’s perimeter. The grains outside the duplicate’s perimeter can then disconnect from neighbouring grains. 

It would be possible by tinkering with this algorithm to create multiple copies of a sample shape, or to scale up to produce a single large copy of an object. 

MIT student and paper author Kyle Gilpin used a car as an example: “Say the tire rod in your car has sheared – you could duct tape it back together, put it into your system and get a new one”.

The paper authors used smart pebbles to test the algorithm, which work as a simplified, 2D version. These cubes have all four faces studded with electropermanent magnets, or materials that can be magnetised or demagnetised with an electric pulse. They can be turned on and off, and the pebbles use magnets to communicate and share power as well as for connecting with one another. Each of the smart pebbles has a microprocessor inside it which can keep 32 kilobytes of program code, and has just two kilobytes of memory. 

To test their algorithm, the researchers designed and built a system of ‘smart pebbles’ — cubes about 10 millimeters to an edge, with processors and magnets built in.
Photo: M. Scott Brauer

To attach to each other, to communicate and to share power, the cubes use ‘electropermanent magnets,’ materials whose magnetism can be switched on and off with jolts of electricity. Each cube has magnets — recognisable by the reddish wires wrapped around them — on four of its six faces.
Photo: M. Scott Brauer 


The researchers said there wasn’t room for more magnets on the cube, however, they did run computer simulations which demonstrated that their algorithm would be able to work with 3D blocks of cubes – by treating the layer of each block as its own 2D grid. In the simulation, cubes discarded from the final shape would disconnect from surrounding cubes. 

At the moment, the researchers are outlining the possibilities of the algorithm. They will present the final paper at the IEEE International Conference on Robotics and Automation this May.