The classic humanoid biped of science fiction and Hollywood movies might look the part but, in reality, those two-legged machines are hard to build.
Getting a bipedal robot to not fall on its face, much less walk, is a feat that no one has mastered.
Roboticists are getting there, though, reports Wired. Take, for instance, a robot called Cassie from Agility Robotics. It’s a highly evolved biped, walking and balancing seemingly with ease. But behind Cassie is a mountain of physics and engineering.
Cassie’s story begins with a different robot named Atrias. Atrias was a research robot, built to help unravel the physics of bipedal walking, and that it certainly did – when its developers walked it across a force plate, its gait dynamics matched those of a human. But it had its limitations: researchers built Atrias to test the scientific concepts of walking, not as a sturdy marvel of engineering, so it had some durability issues.
Cassie, though, is built to last. “It’s the difference between what SpaceX has been doing recently with rockets and what was done with the Apollo program,” says Agility Robotics CEO Damion Shelton. “There was a lot of good science that came out of the Apollo program, but it wasn’t practical or durable as a long-term technological solution.”
Cassie is a highly evolved Atrias, a robot meant to escape the lab and scramble into the market (Cassie is already available for research purposes—Agility Robotics’ larger plan is to turn it into a workhorse for making deliveries and inspecting power plants).
“Cassie can take a fall and survive. It’s a much tougher robot that Atrias was,” says Agility Robotics CTO Jonathan Hurst. “Cassie can steer, pick its direction. Cassie can stand in one place, it has ankles, it can actually balance, whereas Atrias couldn’t.”
The bird-like Cassie (from “cassowary,” the giant bird from Papua New Guinea) is also a radically different looking robot than Atrias, whose springy legs were shaped like diamonds. But while Atrias was able to nail a human-like gait, it wasn’t as efficient as a biped could be. So Hurst took what he learned from the research robot and ran the maths, landing on a limb that just so happens to look decidedly avian.
“This is kind of reassuring and promising to us,” he says. “Maybe we’re starting to understand some of the reasons behind why animal legs are shaped the way they are.”
It’s robotic biomimicry kind of by accident, chasing the inherent efficiency and stability of organisms without necessarily copying them joint for joint and bone for bone. So designing a bipedal bot to work well in human environments isn’t necessarily about designing something that looks exactly human. It’s about building the most effective platform that the laws of physics allow.
At the moment, bipeds struggle at even tackling flat surfaces with any measure of speed. And stairs? There are few better ways to cripple a very expensive robot.
But the beauty and promise of robotics is that there is no single right way for machines to get around. Quadrupeds, like Boston Dynamics’ scary-nimble pack mules, offer great stability. Cockroach bots might one day scramble over rubble and squeeze into tight spaces to inspect collapsed buildings. And bipeds may have their own place in this increasingly robotic world, doing uniquely humans things like climbing ladders and turning valves in decommissioned nuclear facilities, for instance.
So sure, Cassie doesn’t have arms yet. And a human still has to pilot it around. But bipeds are going places.