Robert Shepherd Gives Robots Smooth Moves
“You’re either creeped out by it or excited,” says AAAS member Robert Shepherd.
On a bench at his Cornell University laboratory, a flat, translucent rubber “X” suddenly lifts itself off the counter and begins moving. Uncannily, it undulates like a caterpillar, its body inflating in rolling waves. Suddenly its gait changes to a four-legged walk; it becomes a sort of jaunty, trotting X.
Reaching a plexiglass barrier, it slumps down flat and begins insinuating itself supplely beneath. Once past the barrier, it appears to collapse with fatigue.
Indeed, it is hard work being a robot in Shepherd’s laboratory.
The assistant professor of mechanical and aerospace engineering runs a soft-robotics research lab where no idea is too far-fetched for consideration and innovations fuel fields as diverse as medicine, gaming, manufacturing, defense—and even caregiving.
Soft Robotics is a sub-discipline of robotics in which flexible, compliant materials are used to develop machines that can move fluidly and react to their environment much as living organisms do—a field known as biomimicry. With their ability to morph shape and withstand extremes of heat and pressure, soft robots increasingly are used in manufacturing, food handling, and invasive surgery, among other applications.
But their end-applications are not necessarily what drive Shepherd’s experimentation. “I just am fascinated by animal physiology,” he says. “The technical challenge to us as scientists and engineers is, well, biology can do this … how can we replicate that with our available tools? It’s a puzzle to solve.”
Shepherd’s X—or multigait soft robot, as it’s called—began as a doodle on a napkin. At the time, he was a postdoc in George Whitesides’s laboratory at Harvard, one of the leading soft robotics hubs in the country.
“It was the first demonstration of a robot that could perform both a quadrupedal and undulating gait,” says Shepherd. It created a big stir in the media as well as scientific arenas, but Shepherd is unassuming about it: “For me it was a sandbox to play in.”
Since then, he has put together his own team of highly interdisciplinary undergraduate and graduate student researchers who tackle a range of challenges in organic robotics, from bioinspired machines, soft sensors and displays, to advanced manufacturing.
Although a materials scientist by training, Shepherd brings a knowledge of chemistry to his work, creating products that blend chemistry and mechanical engineering in novel ways.
Among his lab’s most remarkable recent innovations is an electroluminescent silicone skin that can stretch to about 500 percent of its original size and change colors like an octopus responding to external and internal stimuli. Among future applications for this, he imagines portable electronics that can expand from phone to tablet size.
He and his lab also developed a robotic, soft prosthetic hand that can “feel” its surroundings, finely differentiating the shape and texture of objects through internal sensors.
It usually begins with a giant brainstorming session.
When a team member gets “an intuition that we should work on something,” Shepherd says he convenes a meeting to bat around ideas and see if it’s something to explore experimentally. If it passes muster, they “rapidly iterate through some designs, test them, improve, then have another diverging-converging round of meetings until we have something that’s pretty solid that we can write a research paper on.
“I think when you are trying to come up with new ideas the worst thing to have is people in the room who provide a dark cloud of doubt around things,” observes Shepherd. “We want the research we do to be joyful and we also want it to be useful.”
Most of the work begins with liquids, he says. “We use soft materials, manufactured rubber precursors … which makes it a lot easier to make sophisticated functions using manufacturing techniques like injection molding or 3-D printing.” Among the rubbers his team has developed are those that can emit light, sense touch, and stretch.
Many of his robots are balloon-based, meaning they are inflatable machines that are controlled by sending pressured gas through a series of pneumatic tubes, which effectively operate as soft muscles. He engineers these balloons with specific characteristics—where they bend, how they inflate, what force they will apply when they inflate, and the rate at which they will inflate. These determine how and when the robot will move and respond to its environment.
Since his early, tethered version of the multigait soft robot, Shepherd developed an autonomous version with embedded actuators, sensory capabilities and algorithms to respond to different sensory input. “The computation is done off-board, but it is still untethered so we now have a true robot,” he says.
It is perhaps the price of being a pioneer, but Shepherd often finds himself waiting for the “real” world to catch up with his innovations. One example is a haptic-glove interface for virtual-reality gaming he developed that was a big hit at Siggraph. The controller changes its shape based on what gamers are seeing in the virtual reality space, simulating the experience of touch.
“The VR market isn’t growing as fast as people think it is,” he says. “But that doesn’t mean there aren’t niche markets within the VR space.”
His team quickly pivoted and is adapting their technologies for a slew of assistive devices: haptic gloves for surgical training; orthotic gloves that can be worn to augment force in grasping, helping people with strokes; even products in the arena of human caregiving.
On the latter he is vague:
“I can’t talk about that right now because we do try to commercialize some of our technology,” he says, adding: “But I think the therapeutic potential is there. You could do a lot of things with soft robots. Caregivers in hospice especially—there’s only so much time people can spend with their patients and the rest of the time they may wish for company other than a television.
“The control sophistication to replicate human interaction is not there yet … but we may be able to make something that feels more human, while not experiencing the uncanny issue,” he says.
For all of his study of bioinspired machines, Shepherd says his greatest teacher may be his 9-month-old daughter, who wants to walk before she crawls.
He says, “If it’s so hard for humans to learn after billions of years of evolution from bacteria, then it’s not surprising that it’s difficult for robots as well.”