How to run on rough terrain — a guide for humans and robots

A diagram demonstrating the process of “swing-leg retraction” on uneven terrain.
A diagram demonstrating the process of “swing-leg retraction” on uneven terrain.

Running on rough terrain is an incredibly complex process, yet animals from guinea fowl to cheetahs manage to do it. Among humans, even casual runners can navigate such surfaces fairly well.

How they do this has been an ongoing question in the field of biomechanics. A study published March 12 in the journal Royal Society Open Science offers some answers, as well as some guidance for anyone building a robot designed to run. Using mathematical analysis and computer simulations, researchers in the Yale lab of Madhusudhan Venkadesan, assistant professor of mechanical engineering & materials science, found that how your foot hits the ground plays a major role in stability. Also, it’s the changes in the slope of the terrain — not in height — that cause the most problems.

We found that rough terrain running is intrinsically unstable, leading to tumbling falls,” Venkadesan said. “But our analyses show that minimizing tangential, scuffing collisions of the foot with the ground nearly completely alleviate the risk of tumbling.

That is, all that sliding and scuffing that tennis players do on flat courts may work for them, but it hinders anyone going for a jog in the woods. Even if they’re not aware of the mechanics of their motions, most animals learn this quickly. With frame-by-frame analysis of videos, previous research has shown that running animals will retract their leg backward just before the foot hits the ground. As a result, the foot is moving at less speed than the rest of the runner’s body. That minimizes the horizontal collision. Venkadesan’s current research shows that this prevalent feature of running, known as “swing-leg retraction,” is an effective way to maintain stability.

This has the effect of reducing the speed of the colliding foot, and thereby reducing the scuffing of the foot on the ground,” said Venkadesan, who collaborated on the work with Shreyas Mandre, assistant professor of engineering at Brown University.

This proves true for bipeds like humans, as well as adept four-legged runners, such as dogs and horses. And it’s a particularly good thing to know if you’re designing a robot to run.

It’s part of the things you need to pay attention to — it’s an important parameter for your robot,” said Nihav Dhawale, first author of the paper and a visiting research assistant in Venkadesan’s lab.

Robots don’t have the intuition of biological runners, but they have one advantage. Their reaction times to variations in terrain is about 1 millisecond, while it’s close to 100 milliseconds for humans. It’s a critical difference, since the time between steps in running is about 200 milliseconds.

“So while a human has only two chances to correct itself, a robot has 200,” Venkadesan said.


Media Contact

William Weir:, 203-432-0105