At MIT, a student takes a simplistic approach to a complicated concept
At MIT, a student takes a simplistic approach to a complicated concept
November 2013 - Look through a child’s toys and there’s a good chance you’ll find Legos strewn about. Snapped together, the tiny blocks let kids create little worlds using their imaginations. It was while playing with these modular toys that then undergrad John Romanishin, now assistant research scientist at MIT Computer Science and Artificial Intelligence Laboratory, first started thinking about modular robots.
“I’ve always been interested in things that stick together—like in the movies where robots rebuild themselves,” Romanishin says. “There’s a lot of cool stuff out there, but the systems are really complex. There’s not a whole lot of simple and effective options.”
Romanishin had the idea for these simple, modular robots for a long time before deciding to approach MIT Professor Daniela Rus. To show her the idea was something worth pursuing, Romanishin applied for a grant application for an independent summer project at MIT. “So I built a quick prototype, showed it to Professor Rus and she wrote a letter to support the grant application,” he says.
Romanishin got the grant and spent the summer working on the project. “I spent a lot of time making things that didn’t work before I got anywhere,” he says.
His efforts paid off, leading to momentum-driven, magnetic modular robots, or M-Blocks. While Romanishin is certainly not the first to broach the subject of modular robots, his simple design is unique. “Just Google self-configuring modular robots—pretty much every one I’ve seen is very complex, needing multiple motors, and takes a long time to move from one position to another,” he says.
By contrast, M-Blocks have one primary motor as well as a servo to actuate the belt. “Instead of needing a motor for each face of the cube, we have just one primary motor and a servo motor,” Romanishin says. “If the actuation to move a cube acts locally on each of the six faces, instead of acting at the center of mass like our system, there will be more actuators and more complexity. Ours has a really high impulse of torque because of the way the brake is structured. We can use permanent magnetics to hold them together so when it moves from one place to the next, it breaks and moves another magnetic bond but it happens nearly instantly. You don’t need sensing or computation to do that. Other systems use permanent magnets to connect but the problem with those is often times they’re very powerful and it wastes energy to break them apart, requiring a special latch to break them and suddenly you need a bunch of these latches to break connections apart—it’s a very big challenge.”
Breakdown
Each cube is 50 mm. and is bonded through permanent magnets embedded in each edge with additional magnets located on each face of the cube to help align them. The modules move by pivoting around their 12 edges, allowing for multiple movements, as opposed to only moving in one direction on a plane. These movements are driven by a torque generated by rapidly decelerating an internal flywheel, according to Romanishin.
While years away from serving a practical purpose, the team at CSAIL is hard at work. The original M-Blocks were milled out of a single piece of 7075 aluminum. “I love milling. We fit the magnets into these little cages on the frame,” Romanishin says. “We are going to try making the cubes using injection molds with plastic instead of metal because it’s cheaper using metal components inside each cube including the frame that holds the motor and flywheel in place.”
A mechanical engineer by background, Romanishin says there are plenty of people involved in research focused purely on software and algorithms that could one day help M-Blocks think for themselves. “If we can keep M-Blocks roughly the same size or a bit smaller, forming much stronger connections, they could theoretically weld themselves together, also allowing them to form scaffolding or molds for use in the construction industry, for example—specifically designed for whatever job is needed,” Romanishin says. “Maybe the scaffolding will problematically build itself in real time where people are for that day’s job and then can scale itself down at the end of the day, ready to build itself up for a completely different job the following day.”
Funding is a major obstacle to moving the research forward and also determines the pace at which scientists are able to advance the technology. “We’re definitely still in the beginning stages—M-Blocks aren’t anywhere near where they need to be to build scaffolding, for example—yet,” he says. Romanishin says he’d first like the M-Blocks to configure themselves into furniture. “Maybe a cool chair or rebuildable couch.
“Of course, a self-building structure on the battlefield could also be of good use, too,” he says. However, U.S. military programs such as Defense Advanced Research Projects Agency ended funding for a specific grant that was supporting much of the self-reconfigurable robotics research available. “They don’t seem to be focusing on it right now but that could change in the future, which would be great since much of the work done in this field was due to this grant,” he says.
“In the future, the metals and manufacturing industries could use this technology to help set up needed tooling in a factory, making things reconfigurable—there’s all kinds of potential,” Romanishin says. “M-Blocks could be useful to transport tools like for up-and-go mining operations. We just have to dream big to get there.” FFJ
