A soft robotic hand has finally achieved a historic accomplishment: beating the first level of Super Mario Bros. Although quickly pressing and releasing the buttons and directional pad on a Nintendo Entertainment System controller is a fun test of this three-fingered machine’s performance, the real breakthrough is not what it does—but how it was created.
The Mario-playing hand, as well as two turtlelike “soft robots” described in the same recent Science Advances paper, were each 3-D-printed in a single process that only took three to eight hours. “Every one of those robots in this paper was 100 percent no-assembly-required-printed,” says co-author Ryan Sochol, an assistant professor of mechanical engineering at the University of Maryland.
One-step production would make it easier for researchers to develop increasingly complex soft robots. These bots’ squishy makeup lets them interact with delicate materials—such as tissues in a human body—without the kind of damage more rigid machines might cause. This makes them good candidates for tasks such as performing surgery or search and rescue and even sorting fruit or other easily damaged items. But so far most such bots still include at least some rigid components. It was not until 2016 that researchers created one entirely from flexible materials. To make that octopuslike soft robot work, its creators had to ditch rigid electronic circuits for a microfluidic one. In such circuits, water or air moves through microchannels; its flow is modified by fluid-based analogues to electronic components such as transistors and diodes.
In the new study, the researchers stepped up the development of this technology. “They introduced much more complicated microfluidic circuits,” says Harvard University engineering professor Jennifer Lewis, who co-authored the 2016 paper but was not involved in the University of Maryland’s project. In the Mario-playing hand, for example, the circuit allowed a single source of fluid to send different signals, telling each finger to move independently by simply varying the input pressure.
Printing It Up
But in making soft robots more sophisticated, fluidic circuits also render the machines harder to manufacture and assemble. That is why Sochol is so excited about printing them in one step. “Never once has it been done all in a single run,” he says, “to have an entire soft robot with all of the integrated fluidic circuitry and the body features and the soft actuators [moving parts] all printed.”
He and his colleagues used a PolyJet 3-D printer, a type that sets down a liquid layer, exposes it to a light that solidifies it and then adds the next layer. The model they employed, manufactured by a company called Stratasys, could produce three types of material: a soft rubberlike substance, a more rigid plasticlike one and a water-soluble “sacrificial material” that acts as scaffolding during printing but must be removed from the final product afterward.
Such high-tech printers can retail for tens of thousands of dollars—but Sochol’s team did not need to buy one. “We use a service on campus to do this,” he says. “So we sent our files to them, they printed it, and then we picked it up.” Sochol estimates that anyone else wanting to print one of these designs—which his team shared as open-access software on the development site GitHub—could use a similar 3-D-printing service for about $100 or less.
Sochol contends this process is faster, cheaper and easier than fabricating a microfluidic circuit in a clean room, creating a robot separately and then combining them later. Lewis does not entirely agree. “There’s an elegance to it. I’m not sure it’s faster, cheaper, necessarily,” she says. “But there is cumbersome nature to having to create the circuit by one method and then insert it, like we did, into a molded and 3-D-printed robot. And I would say that the method that [Sochol and his colleagues] chose ... has many advantages in terms of being able to print multiple materials of different stiffness.” Lewis also points out that the new soft bot is not ready to go immediately after printing. “One cumbersome part of their method is that you have to remove all the sacrificial material,” she says. “And when it’s on the outside of the body, just as support, that’s fine. But it’s also present in all of those internal channels.”
It’s-a Me, Mario!
After cleaning up their printed robots, Sochol’s team had to design a performance test. Earlier studies have programmed robotic fingers to play a tune on a piano, for example, but Sochol’s team thought that task was too easy. “With that, we could set the tempo arbitrarily, he says. “If the robot misses a note or something like that, there’s no meaningful penalties.” Video games seemed a little more uncompromising. “If you make a mistake, if we don’t press the button at the right time or we don’t [release] the button at the right time, you can run into an enemy, you can fall down a pit, and it’s an immediate game over,” Sochol says.
The researchers placed their three-fingered robotic hand on a Nintendo controller, with each finger laid on a different button or the directional pad. By feeding fluid through a control line at different pressures, they could make each finger respond. “For a low pressure, the circuit is able to respond to that and press only the button that causes Mario to move forward,” Sochol explains. “And then for a medium pressure, a second finger begins to press a button, and now Mario can run. And then if it’s a high pressure, then all three fingers will be pressing their respective buttons, and Mario will jump.”
The team wrote a computer program that would change the pressure automatically, causing the fingers to move in a set pattern. Because people have been playing Super Mario Bros. for decades, the team knew exactly what sequence of buttons the hand would need to press to win the game’s first level. It just had to run through that preprogrammed list with the correct timing—which is harder than it sounds. The challenging part, Sochol says, was “getting it to not just press a button but then stop pressing it and then repress it, because there’s a lot of times where Mario has to jump and then jump again very quickly as he’s running.”
“The Mario part is kind of cute and certainly will be attention-grabbing,” Lewis says. “But I think what’s really powerful about this paper is the multimaterial 3-D printing, the ability to integrate all of this complex fluidic circuitry in one fab step. There’s really a lot to like about what [the researchers have] done.”
Winning the video game showed that the fully printed robotic hand could respond swiftly and accurately to a changing input. Any well-known video game could have made this point, but Mario holds a special place in many players’ hearts. “We felt like this was the baseline game,” Sochol says. “When I was a four- or five-year-old, and we got a Nintendo system, that was the very first thing that I played ever.”