Christmas came early this year for robotics and technology enthusiasts everywhere, thanks to Boston Dynamics publishing a video of their latest creation. PETMAN (Protection Ensemble Test Mannequin) won’t be completed until sometime next year, but it’s already drawing comparisons to science-fiction robots like the Terminator. Designed to test chemical protection suits for the U.S. Army, PETMAN can walk, run, crawl, and perform other physical stress tests. Eventually, the robot will be covered in a synthetic skin that simulates human perspiration in order to test how the suits deal with moisture internally. Once it’s wearing a hazmat suit, it may be difficult to tell that it’s a robot at all.
What makes PETMAN so revolutionary has actually been in the works for more than two decades. Marc Raibert, the founder of Boston Dynamics, made a name for himself in the ’80s at MIT’s Legged Robotics Lab, where he developed walking and hopping robots with unprecedented balancing capabilities. A few years ago his company’s DARPA-funded BigDog, a four-legged robot designed to haul hundreds of pounds of military gear, became an internet sensation. A video showing the robot tackling difficult terrain with lifelike movement, including mud and ice – places legged robots simply hadn’t gone before – went viral overnight. Now the same approach has been successfully implemented in a bipedal robot for the first time, where many others (including Honda’s ASIMO) have relied on a different, more limited approach entirely.
For more than 20 years humanoid robots developed at Japan’s Waseda University and then Honda have pioneered an approach based on the zero-moment point, or ZMP. It allows a robot to maintain its balance by keeping its center of gravity within set boundaries at any given point in the robot’s walking gait. The result has been robots which are able to walk upright without falling over, but they generally move much slower and less naturally than a person. Honda’s ASIMO is an engineering marvel, but after more than 20 years of research and (according to some sources) upwards of $60 million dollars invested, it can barely run faster than PETMAN can walk. And “running” is a term used loosely, given that ASIMO’s feet are only airborne for a fraction of a second. Shove ASIMO roughly and it falls over, whereas PETMAN stumbles momentarily before it finds its footing and keeps going.
Video (MIT’s Leg Lab):
Honda will still have some technical details to hold over Boston Dynamics even when PETMAN is complete: energy autonomy, for instance, but it seems ASIMO’s title of “world’s most advanced humanoid” is in jeopardy. ASIMO’s battery pack and self-charging station, as well as its onboard computer, allow it to go where PETMAN can’t. PETMAN will remain tethered to an external power supply and computer processor, which will be piped in through its boot (to avoid puncturing the hazmat suits it will wear). And ASIMO already has some rudimentary artificial intelligence which PETMAN won’t need to do its job; stuff like face recognition. Nor will it have to manipulate objects or navigate its environment. But with strong artificial intelligence still just a dream, the current state of the art in humanoids is measured not by brains but by how closely they replicate human movement, and it’s clear PETMAN does it better.
I’m not saying that the ZMP method is obsolete; it’s a proven way to move around structured environments like houses, hospitals, and office buildings. However, just as Honda’s ASIMO and its prototypes stunned the world beginning in the late ’90s, inspiring dozens of copies across Asia, I predict that PETMAN will have a similar effect. And ultimately, it will bring about a more robust generation of humanoid robots.
[source: Boston Dynamics @ YouTube]