Switzerland's ANYmal looks to be emerging competition for Boston Dynamics' Spot.
While there’s been a huge amount of innovation in humanoid robots from the likes of Figure, Apptronik, Sanctuary AI, and the Tesla Optimus bot, there’s still room for other form factors. Boston Dynamics has Spot, a quadrupedal sensing and inspection robot, for instance. ETH, the Swiss institute of technology in Zurich, has been working on ANYmal for some time. And it just got smarter and more capable: no longer just hiking the mountainous trails of Switzerland, ANYmal is now doing parkour.
“[ANYmal] is proving rather adept at parkour, a sport based on using athletic manoeuvres to smoothly negotiate obstacles in an urban environment,” says ETH. “ANYmal is also proficient at dealing with the tricky terrain commonly found on building sites or in disaster areas.”
The new capability is thanks to AI, but not only AI.
Researchers and doctoral students at ETH used machine learning techniques to allow ANYmal to teach itself. To learn, in other words, like a child via trial and error. The result is that now when the four-legged robot approaches an obstacle, it senses the terrain with its camera and then applies its onboard artificial neural network to classify the obstacles and decide on an approach.
But there’s more than machine learning at work, researchers say. They’ve also given ANYmal a cheat sheet or a head start via model-based control.
Together, both approaches achieve optimal results.
“This provides an easier way of teaching the robot accurate manoeuvres, such as how to recognise and get past gaps and recesses in piles of rubble,” ETH said in a press release. “In turn, machine learning helps the robot master movement patterns that it can then flexibly apply in unexpected situations.”
It remains to be seen whether this teach yourself plus we’ll-teach-you approach is the best overall approach. But it does model precisely how humans learn: some trial and error, and some education and training.
ANYmal is commercially available via ANYbotics, a spin-off company from the university which secured a $50 million funding round last year. The company markets ANYmal as an autonomous inspection robot for difficult and dangerous environments, mounting it with visual and thermal cameras as well as LIDAR. It can carry a 10 kilogram payload (about 22 pounds) plus additional sensors, and runs on battery power for 90 minutes.
Competitors beside Boston Dynamics include KNR Systems, Lymxmotion, Moog, and Unitree. Uses include industrial, search and rescue, and military, and telemedicine.
It’s not clear how thriving the market is—Boston Dynamics only says it has “over 1,000 robots in customer hands today—though analysts forecast significant growth, up to 8.8% year-over-year, for the next decade.
