Kevin Albert is playing the part of lion tamer. The 33-year-old engineer, with a passing resemblance to Joe Namath, sticks his head into the four-fingered grip of a robotic hand, pauses for a moment and then emerges unscathed.
This is no ordinary robot: It’s made of fabric and air rather than steel and motors. Albert, co-founder of the San Francisco startup Pneubotics, is demonstrating how a new generation of robots can be made safer and more versatile — allowing them to move off the factory floor and work in close proximity with humans.
Though the robot population has risen to 1.6 million since General Motors Co. first put one on an assembly line in 1961, growth has been limited because most are variations on the original theme, dangerous claw attached to metal limb. Companies like Pneubotics are creating alternatives that, if they work, could bring the productivity of robots to more industries, including construction, warehousing and agriculture.
“What we are trying to do is open the space of what robots can get to,” says Albert, who previously helped develop quadruped robots at Boston Dynamics Inc., now part of Google Inc.
Albert’s co-founder at Pneubotics is Saul Griffith, a 41-year-old engineer whose inventions earned him a $625,000 MacArthur genius grant in 2007. Their office, a former pipe-organ factory in the Mission neighborhood, looks more like a software startup than a robot engineering workshop.
In another demonstration, Albert has a red tentacle nicknamed Elephant Trunk wrap around him. It hisses and crackles like a beach toy as it’s inflated.
The arm has neither joints nor a skeleton and is mostly made of heavy-duty nylon cloth. The main shaft is surrounded by chambers that curve when inflated, making the arm bend. When deflated, it can fit in a sock drawer.
“If you really want the traditional robot to be useful, like being able to lift things, it has to be strong,” said Morten Paulsen, who covers industrial automation giant Fanuc Corp. as head of Japan research at CLSA Asia Pacific Markets. “Once it’s strong, it is essentially dangerous.”
Rigid robots require a space free of obstacles, known as the envelope, to avoid harming humans. Inflatables, by contrast, can conform to the environment. To demonstrate, Albert maneuvers Elephant Trunk head-on into a steel pillar in his office; the arm buckles, twists, then squeezes by and regains its shape.
Several other companies are developing collaborative robots, machines that can work alongside humans, including Fanuc, ABB Ltd. and Kuka AG. Rigid robots can be made safe by adding so-called compliance, or the ability to give way against resistance, with complex specialty motors and expensive sensors. For soft robots, compliance is built in.
Pound-for-pound, traditional robots are also surprisingly weak. Fanuc’s M-2000iA can heft 1.2 tons, handling car chassis and tractor frames with the ease of a baton twirler. The robot itself clocks in at 8.6 tons.
The strength-to-weight ratio for conventional robots is 5 to 10 times, meaning it can take 80 pounds (36 kg) of metal to move one-gallon of water (about 8 pounds). An inflatable arm designed by Pneubotics can lift more than five times its own weight.
This combination of safety and strength can open swaths of the economy to automation’s benefits.
Robots helped more than double labor productivity of U.S. manufacturing in the 20 years to 2010, according to Bureau of Labor Statistics data. At the same time, output per-hour grew only about 35 percent in transportation and warehousing, 22 percent in services, while declining for construction — all areas where robots have been too dangerous for ordinary use.
Compliant materials have another benefit: they take greater advantage of computing power. The basic math needed to animate a rigid system can be worked out on the back of a napkin. Inflatable bodies use control algorithms that calculate gas dynamics and material strain 1,000 times a second. Their physical abilities are more closely hitched to the exponential drop in the price of microprocessors, known as Moore’s Law.
“The moving parts of rigid machines don’t get the advantage of Moore’s Law because they are generating forces and not information,” said Christopher Atkeson, a professor at Carnegie Mellon University. Atkeson’s research into inflatable robots for health care inspired the huggable protagonist in Walt Disney Co.’s animated film “Big Hero 6.”
Griffith has a record of innovating by using computing to achieve what was previously done by mechanical means. His quest has been for solutions that reduce complexity, weight and cost.
Makani Power, a company he helped found to improve wind power efficiency, replaced the windmill structure with a kite on a tether. Such a design, made possible by intensive flight-control computation, generates electricity more efficiently. Another benefit: a 90 percent reduction in materials.
Griffith’s foray into robotics started innocently enough with a large blow-up elephant — a hand-made Christmas present for his 6-year-old niece. When she got bored with the toy, Griffith found himself looking for ways to make it walk.
He wrote a program to control how airflow expanded and contracted in pockets around the elephant’s knees, making it totter forward. The child wasn’t the only one impressed — the Defense Advanced Research Projects Agency gave Griffith a grant as part of its Maximum Mobility and Manipulation program.
The next step was a six-legged inflatable beast capable of carrying two adults on its back. A mix between an ant and a cockroach, it was nicknamed Ant-roach. (The better of two possible names, Griffith joked.)
Pneubotics is now ready to make a machine capable of commercially useful tasks. The company plans to deliver a low- cost lifting robot for logistics, manufacturing and medical spaces as soon as a year from now, Albert said.
“Even if electromechanical robots were twice as good as they are today, they still wouldn’t be good enough. And we don’t even know how to make them twice as good,” said Griffith. “You have to change the game.”