MIT Robotics Team Uses Self-Sculpting Sand to Duplicate Objects
It’s clear that at the Massachusetts Institute of Technology (MIT), recycling has taken on a whole new meaning. Specifically, researchers at MIT’s Distributed Robotics Laboratory (DRL), at the university’s Computer Science and Artificial Intelligence Laboratory, are forming a new way to simply remake any object that already exists by inserting that object into what’s called “self-sculpting sand.”
What is self-sculpting sand? We’ll try our best to explain that in a minute. But for now, what the MIT scientists and researchers have done on a limited basis, and hope to do on a much larger basis in the future, is develop a process that not only will allow for immediate duplication of a tool, but also allow for the possibility that that object can be turned into something else.
That’s right: A hammer could be made into a wrench. A tiny footstool could, seconds later, produce a full-size footstool, ready to be used so you could reach the top shelf where all the goodies are hidden in our kitchen.
It sounds like something out of The Jetsons, but according to Kyle Gilpin, who along with MIT Professor Daniela Rus will present a paper on the self-sculpting sand research at the IEEE International Conference on Robotics and Automation in May, self-sculpting sand is a reality that could change the world.
“It’s obviously just one form of a recycling system, because you’re not just using something once,” Gilpin told me in an interview last week. “You’re making an object into another object that you can then use. It’s revolutionary work, and it’s not that far off in the future from becoming a reality.”
Gilpin, a Ph.D candidate in the Electrical Engineering and Computer Science Department at MIT, is one of about 20 researchers working specifically on the self-sculpting sand project. He said he originally came to MIT as an undergraduate studying electrical engineering and computer science, and then wanted to work on his doctorate in something else.
That is how he stumbled into the Robotics lab, where, in addition to the self-sculpting sand project, there are projects involving flying robots, 3-D robots (“sometimes we have to duck while walking through the lab,” Gilpin joked) and other high-tech research situations.
Before getting into the details of how self-sculpting sand works, Gilpin first gave me some background on how the project came to be.
First, Gilpin explained, Professor Rus at MIT was looking for a modular robotic system that could self-disassemble. What she wanted to do was to find a system that can use shapes to repeat the process of making objects.
That involves building miniature modules, smaller and smaller, Gilpin said. But creating so many modules makes what the long-term goal of the project at the start — modules that could, in effect, “talk” to each other — difficult.
“As you miniaturize modules and they get smaller and smaller, you need hundreds of thousands of individual modules, because you’re making the parts smaller,” according to Gilpin. “Once you have 1 million modules, you need some way to communicate with them to make them efficient. We need a way to tell all the modules in the object you’re forming what to do, like you would tell a robot what to do.
“It’s just not possible to tell a million modules what to do,” he continued. “You can’t tell them ”you do this, you do this’ — there’s too much data that has to be communicated.”
To try to solve that problem, Rus, Gilpin and the MIT team had to figure out a way to have the modules tell themselves what to do.
They did this by first creating smart sand pebbles, which Gilpin called “robot pebbles.” These cubes have four faces that are stacked with “electropermanent magnets,” materials that are magnetized or demagnetized with a single pulse. These are different from the usual permanent magnets we see today, as these magnets can be turned on and off like a switch; they don’t need a constant current to stay magnetized.
The pebbles then use the magnets to not only connect with each other, but also to communicate and to share power.
Gilpin explained that each pebble additionally contains a tiny microprocessor, which can store 32 KB of program code, along with 2 KB of working memory.
The pebbles have magnets on only four faces, Gilpin explained, because with the addition of the microprocessor and circuitry to regulate power, “there just wasn’t room for two more magnets.”
The robot pebbles are then contained in a flexible circuit board that is wrapped around a brass frame for use with the next part of the project.
Once the pebbles have been created, the process of duplicating can begin, Gilpin explained:
We take a model whose shape we want to form, like a miniature hammer. And we dip into this bag of smart sand. The modules in the bag sense the shape of the prototype object, and then they go about forming a duplicate of that original hammer. It can be a magnified version or a version that’s the same size.
The pebbles are not pre-programmed with the shape of the hammer. They have a more generic program that helps them learn how to make the hammer. It’s involved in sensing and duplicating the border of the original object. [The pebbles] create a duplicate of the border, and then there’s a flood process, and anything inside that duplicate border is replicated.
Basically, the pebbles are passing messages back and forth to each other, telling the other pebbles what is the shape of the object they’re going to make. Through this process, a hammer can be made into a larger hammer, or a wrench, or a screwdriver.
All of this was theory for the MIT team until about a year ago, when the first tests were done. They were able to replicate miniature hammers, Gilpin said, and performed successful tests upwards of 100 times. “In simulation, we’ve done it successfully thousands of times,” Gilpin said.
The MIT Robotics lab isn’t the only university team working on self-sculpting sand projects, Gilpin noted, pointing to groups at Carnegie-Mellon University and the University of Pennsylvania.
Thinking about the real-world possibilities of self-sculpting sand gets the mind excited. After speaking with Gilpin, I spent quite a bit of time just imagining the possibilities. Think about how much money the average person would save, for example, on tools. Or on household objects they weren’t using anymore, that suddenly could be made into more useful objects. That old ashtray your kid made in shop class when they were 8? How about turning it into a small night table, for example.
Gilpin admitted that the system the MIT team has come up with “isn’t something that’s commercially viable; but maybe in 5-10 years we’ll be getting closer — we’ll see incremental improvement every year.”
“With any new technology, it’s going to be expensive in price at first, but we hope that it’ll be cheaper by hand,” Gilpin said. “We’re going to have to go to some kind of mass production process [of the smart sand robot pebbles].”
In the batch of 30 modules the MIT team made, each one cost around $400, including the fabrication of the circuit board and the fabrication of the modules.
Still, getting this far has been exciting for the whole team. Gilpin said he hopes the work the team is doing will inspire young people, much as he was inspired as a kid.
“I remember being 10 or 11 and watching ‘Scientific American’ on TV and being really fascinated by that, and having my interest in science sparked,” Gilpin said. “I hope the work we’re doing has that effect on kids today.”