Robots' motor skills are pretty much on par with how I feel while in line for a McFlurry at 4 AM after having a bunch of drinks. Walking is easy peasy, running is a little tricker, and as for jumping around, well, I'd rather not try my luck.
But SALTO, a little robot from PhD candidate Duncan Haldane at the University of California Berkeley, is looking to change that. SALTO can't run or even see, but it can jump higher and faster than any robot before, and even higher than a human. According to Haldane, SALTO is the first step toward creating a robot that can jump around uneven urban environments, just like a person doing parkour.
SALTO can jump nearly two metres per second, and even more impressively, launch off of a wall for a second, higher jump. This might seem like behaviour most suited to video game characters, but SALTO's impressive ability is inspired by nature—to be exact, the adorable little mammalian bush babies found in rainforests. Bush babies use a crouched position to store energy in their powerful legs. Then they chain a series of jumps together to reach far distances.
"The [bush baby] is the proof of concept—there's a system that can do this kind of jumping that exists in the universe," said Haldane. "That gave us the confidence to proceed and make a robot that can do what other robots can't."
The secret to the diminutive SALTO's ability is building this concept into a deceivingly simple one-leg design. Instead of a powerful motor, the robot stores energy in a physical spring at the beginning of its first jump and then releases it all at once during the second jump off of a wall. Animal models have been used to build better robots before; for example, MIT used a cheetah as a model for a running bot.
Because it's a technique that relies on clever engineering rather than super-advanced technology, Haldane said, it can easily be translated to larger robots.
"It's a general truism of robotics that the larger it is, the more expensive it's going to be," Haldane said. "But the principle we pulled out of biology applies across all scales. The jump performance should be maintained with a larger system."
The end goal of Haldane's work on SALTO isn't just to produce a robot that can parkour with the best of them, but one that can help people, too. A robot that can jump across uneven terrain would be hugely useful during, say, disaster relief operations.
But let's be real, a robot that can do parkour is very rad.