These Magnetic Robots Can Selectively Assemble Themselves to Achieve Complex Forms

The demonstration video for this project makes a profound point: almost everything that humans create is built from the top-down — we use big things (both literal and in terms of investment) to make smaller things. But nature does the opposite by using small, basic building blocks (cells) to form larger organisms that are more complex than the sum of their parts. Nature’s approach is far more efficient and offers a huge amount of flexibility. In a quest to replicate nature’s elegance, MIT CSAIL engineers developed these magnetic “robots” that can selectively assemble themselves to achieve complex forms.

Each of these robots is a small cube about an inch to a side. They do not contain any electronic components or means of actuation, which makes the “robot” moniker rather tenuous. That brings the cost for each unit down to a mere 23 cents. Like most biological cells, a single one of these robots isn’t capable of achieving anything. But many working together can become something more. These robots still aren’t capable of interacting with their environment, no matter how many there are. However, they can assemble themselves into pre-programmed (and reprogrammable) arrangements.

They key to this ability is mixel (magnetic pixel) technology, which we covered a couple of months ago. That article even hinted at these robots. Each of a robot’s six faces is a mixel display, which means that it is a magnetic surface that can receive polarization. Importantly, individual points on the surface (the mixels) can receive unique polarization. So the mixel located at coordinates (20,45) can be positive, while the mixel located at (21, 45) can be negative. By making the mixel “image” on each face unique, the team can ensure that a given face on one cube only attaches to a matching face with the inverted mixel image on another cube.

The unique mixel images determine which cubes attach to each other and in what orientation. To force them into interaction, the team put a bunch of cubes into a tub of turbulent water. As the robot cubes tumble around in the water, they eventually find their mates by sheer chance. Given enough time, they all attach as designed to form a complete object.

This is cool, but has limited utility as it is. However, imagine if each cube were smaller — much smaller. A factory could churn on tiny cubes and program them with mixel images. Then an end user could simply dump them into a bottle of water, shake the bottle around, and pull out an assembled object.

More details can be found in the team's paper here.