Why Animals Are Still Faster Than Robots

Why Animals Are Still Faster Than Robots

Despite advances in robotic components, a new study finds that modern robots lack the integration capabilities to support locomotion at the same level as the animals that often inspire their design.

Cat owners know the drill, right down to the butt wiggle.  

Just before a cat pounces, it often shimmies its hindquarters, pressing its back legs further into the ground as a means of gaining more traction to propel itself forward to run toward its prey. These movements are rapid, coordinated, and the byproduct of thousands of years of evolution. And, as a new study in Science Robotics demonstrates, roboticists looking to design a system to mimic such a run would be hard-pressed to do so in a way that could capture the speed and motion of the act in the exact same manner.  

Kaushik Jayaram, a roboticist at the University of Colorado at Boulder, said he and colleagues associated with Robert Full’s laboratory at University of California, Berkeley, started discussing whether animals or robots would prevail in foot races nearly 10 years ago. He said it began as a “bar-type conversation,” eventually leading to the lead authors Samuel Burden at the University of Washington and Maxwell Donelan at Simon Fraser University, to try to come up with a more formal empirical approach during the COVID-19 pandemic. 

“We decided to examine the question by first constraining the problem to two-legged location so we could do a more rigorous analysis—and then split up performance into subsystems to compare animals and robots,” Jayaram said. 

More for You: Punyo: A Leap Forward in Soft Robotics

When the researchers looked at data from dozens of studies, looking specifically at sub-systems including power, frame, actuation, sensing, and control, they found on the whole that biological organisms, whether cockroaches, cheetahs, or human beings, were faster than bio-inspired robotic systems. Even when some of those subsystems performed better than those found in the animals.

Kaushik Jayaram, right, with graduate student Heiko Kabutz, left, in Jayaram's lab on the CU Boulder campus. Photo: Casey Cass/CU Boulder
For example, some of the power sources for robots used high-quality lithium batteries. These can provide up to 10 kilowatts of power for every kilogram of weight. Animal tissue can’t compare—it only produces approximately one-tenth of that. Motors may have more torque than that found in muscles. Yet, when you put all the pieces together, the animals still showed superior performance when it came to locomotion. Jayaram chalked this up to superior system integration. 

“If you think about muscle, it brings all of these subsystems together really well. It’s a frame, an actuator, a controller, power, and sensor all in one,” he said. “Everything is integrated at the cellular level, and we can’t do that yet with our robotics. We are still constrained by trade-offs. We are choosing different components to maximize certain functions and that often means that we are losing some sort of capability somewhere else.” 

The research group initially hypothesized that perhaps they would see differences based on size—maybe a cat-sized robot would be able to beat out a feline in a sprint. But that did not turn out to be the case. 

Become a Member: How to Join ASME 

“We thought maybe we’d see differences in those domains because, on the engineering side, you’re really good at making things at a certain scale with a lot of techniques,” he said. “But that didn’t turn out to be true. While there are challenges at building robots at some of these extreme size scales, it didn’t seem to be a major differentiator when it came to locomotion.” 

While some might see the study’s results as a disappointment, Jayaram said he hopes that it will inspire roboticists to keep working on better integrated systems—and perhaps even new, robotic materials that are more muscle-like in their integration of different subsystem components. No cat butt wiggle required.  

“While it’s true that biology has had generations to come up with its designs for running, we aren’t constrained in the ways—we aren’t limited by our biological ancestors or the way a certain type of biology works,” he said. “We are building up tools, especially with machine learning-based algorithms and computer simulations where we can simulate thousands, even millions, of different design iterations for locomotion systems fairly rapidly. I feel confident that we can potentially surpass animal locomotion, maybe even in my lifetime. I’m quite optimistic that we will be able to realize some promising improvements.” 

Kayt Sukel is a technology writer and author in Houston. 

You are now leaving ASME.org