Robot Design Goes Back to Nature


My obsession with the science of animal motion began with Jerry, my wife’s toy poodle. When Jerry got wet or got something stuck on him, he would twist his spine and shake his entire body at the astonishing rate of four times a second, as I discovered by taking high-speed footage of him. A clothes dryer would take 20 minutes to get rid of as much water as he managed to remove in a few quick shakes.

As a mechanical engineer and biologist, this set me thinking: What if I could design a machine that moves as effectively, flexibly and adaptively as Jerry’s spine? And what lessons might we be able to draw from the extraordinary physiology of other animals?

For millions of years, as animals have evolved to take myriad shapes and forms, they have adapted to solve a variety of physical challenges. Many have overcome obstacles that humans face as well. With the rise of new technologies to measure and analyze their movements, we can now see animals with more clarity and precision than ever before. The research is having a significant impact on robotics, materials science and a range of other fields.

Jerry’s fellow dogs and a number of other species have flexible spines supported by pliable back muscles and controlled by a network of neurons called the central pattern generator; this combination allows them to turn, twist, run, swim and recover from a trip or misstep without the lag time of waiting for commands from the brain. Most modern commercial robots, by contrast, are controlled by a central processing unit that sends commands to each of the joints.

‘Salamandrica Robotica’ in a pool

A decade ago, a team at Switzerland’s Biorobotics Laboratory at EPFL, in Lausanne, built a four-legged walker called Salamandra Robotica. It is able to move autonomously without feedback from a central processing unit because its body segments are controlled by central pattern generators. Upgraded in 2013, it can make a seamless transition from water to land. It’s like an all-terrain wind-up toy.

Snakes provide a different model of spine flexibility as they twist, bend and compress themselves. A snake-like robot developed at Carnegie Mellon University can slither into spaces too narrow for other types of devices. Each of its individual segments senses its environment and calculates how much force to exert and in what direction. One such snakebot was deployed to assist in search and rescue operations in the 2017 earthquake in Mexico City.

A long-standing difficulty for robots is moving through sand, which covers a third of the Earth (including several recent combat zones for the U.S. military) and even more of our nearest planetary neighbors. Because sand is composed of small grains that together can act like a fluid or a solid, depending on the conditions, its subtle physics creates sinkholes that can quickly bury large objects. In 2010, the $400-million Mars rover Spirit ended six years of exploration when its wheels got ensnared in a sand trap. As the rover struggled to free itself, it sank deeper and deeper.

The ‘Sandbot,’ a scaled-down version of an all-terrain RHex robot, seeks to whip its curved legs through porous surfaces in the mode of a sandfish lizard.
The ‘Sandbot,’ a scaled-down version of an all-terrain RHex robot, seeks to whip its curved legs through porous surfaces in the mode of a sandfish lizard. Photo: Georgia Tech; Joel Sartore/National Geographic/Getty Images

But some animals revel in sand. The sandfish, a lizard about the size of a hand, can virtually dive and swim through it, whipping and churning its legs to take advantage of sand’s fluid or solid properties. To replicate some of that motion, researchers at the University of Pennsylvania developed RHex, a robot with six quick-whipping legs, and a scaled-down version dubbed the Sandbot.

While most all-terrain vehicles use large wheels or treads to better distribute their weight and keep the sand solid underneath them, RHex modulates the speed of its distinctive feet to keep the sand from becoming fluid, like a person walking on thin ice. The robot is now being manufactured by Boston Dynamics for military applications, among others.

Robot Design Goes Back to Nature
Photo: Biorobotics Laboratory at EPFL

Videos: See More of the Robots in Motion

Water poses an even more complex problem for robots, and here guidance has come from another animal: the insect known as a water strider, which moves exclusively across the surfaces of bodies of water.

How it does this perplexed biologists for decades. They eventually found part of the answer in the insect’s flexible legs, which are covered with tiny hairs that create air cushions and keep it from breaking the surface. The big puzzle was how striders, especially baby striders, could accelerate to start moving and keep moving fast enough to generate the water ripples that propel them. That question, known as Denny’s Paradox after the Stanford biologist Mark Denny, was solved by scientists (including me) by studying the fluid dynamics just under the surface as the striders moved. It turns out that their air-cushioned legs create tiny dimples on the water surface that gather into complex, churning wakes. These do more than water ripples to propel them.

A robot version of a water strider faces off against the real insect in a Seoul National University lab.
A robot version of a water strider faces off against the real insect in a Seoul National University lab. Photo: Seoul National University

With this research in hand, my colleagues and I at MIT (where I was then a graduate student) were able in 2003 to build the world’s first dry rowboat, the Robostrider. Powered by an elastic band, it skims atop the water on flexible legs that create the robot’s own dimples and vortices. New fabrication technologies have enabled other researchers to design and test at least 20 much-improved versions, most recently at Harvard and Seoul National University. The hope is to use them as cheap devices to deploy widely for surveillance and for chemical monitoring of the oceans, including measuring oil spills.

For movement through the air, animal attributes also have human ingenuity beat, especially when it comes to navigating obstacles. The honeybee picks up 30% of its weight in pollen on every trip to and from the hive, but even more impressively, it does this while brushing past hundreds of plant stems waving back and forth in the wind.

More essays

A standard human approach would be to design a miniature airplane, with a complex camera and computer, to dodge each plant stem. But nature’s solution is simpler and more effective: a resilient bee that survives running into flowers more than 400,000 times in its lifetime.

The bee can do this because of an elastic material in its wings called resilin, which allows it to crumple its wings like folding origami and bounce back. Drone designs are now being developed at EPFL and elsewhere using bee models—not replicating their wings but using springs and detachable magnetic parts to make them more collision-proof.

EPFL built a ‘Cheetah-cub’ with spring-loaded legs to mimic the gait of its namesake at a house cat’s size.
EPFL built a ‘Cheetah-cub’ with spring-loaded legs to mimic the gait of its namesake at a house cat’s size. Photo: iStock; Biorobotics Laboratory. EPFL. Switzerland

The flexible spine and gait of dogs and other mammals have influenced a growing menagerie of robot designs. EPFL has built a catlike “Cheetah-cub” with spring-loaded legs that weighs just over two pounds and what it calls a Bobcat to copy the distinctive bounding movement of the real thing. Boston Dynamics says that its Wildcat, which runs at nearly 20 miles per hour, is the fastest-moving robot yet. Starting next year the company plans to market its Great Dane-sized SpotMini to businesses as a dog-like helper to fetch and carry items around workplaces.

Until now, robots have mostly been used in factories, where they perform repetitive motions like installing windshields or applying paint. For robots to be able to go out into nature and other settings, they have to be prepared to navigate a range of environments and recover from multiple mistakes. Studying sandfish, water striders and bees probably won’t transform the way that humans travel the world, but it may revolutionize the movements of our machines. Robots are becoming increasingly lightweight, autonomous and mobile, and soon they may become as agile as a toy poodle shaking itself dry.

Appeared in the November 10, 2018, print edition.

Leave a Reply