The bipedal hominids may come, but the quadrupeds are already here. They sit in laboratories, inspect power plants and refineries, play football, and even—much to the concern of many—become police officers.
Spot from Boston Dynamics is easily the most recognizable of the bunch, but many startups and research institutions have put their own spin on the category. Heck, even Xiaomi made one for some reason. While biped suppliers want to prove their work, quadrupeds get the job done.
The team at Google’s DeepMind (which recently swallowed up much of Alphabet’s beleaguered Everyday Robots team) issued a research document outline a potential benchmarking system to quantify the performance of these machines. With a name like “Barkour,” one has to wonder if the department has worked backwards from the title.
Google Research points to the various impressive feats quadrupeds have accomplished over the years, from rock climbing to running and jumping (“flipping is much easier than walking,” an MIT professor once told me), but there isn’t really a baseline for determining system efficiency.
Since these machines are inspired by animals, the research team determined that real animals would provide the best analog performance over their robotic counterparts. That meant setting up an obstacle course in the lab and letting a dog run – look at the tenacious little wiener above. The course consisted of four obstacles in a 5×5 meter area, which it notes is denser than the dog shows that inspired it.
Performance is rated on a scale of 0 to 1 – a simple binary number to determine if the robot can successfully cross space in the roughly 10 seconds it takes a similarly sized dog to do so. There are different penalties for slow speeds and skipping or failing obstacles on the track. Google concludes:
We believe that developing a benchmark for legged robotics is an important first step in quantifying progress in animal-level agility. (…) Our findings show that Barkour is a challenging benchmark that is easily adaptable, and that our learning-based method for solving the benchmark provides a quadruped with a single low-level policy that supports a variety of agile low level skills.
The organization says Barkour has proven to be an effective benchmark, even in the face of the inevitable unexpected events and hardware issues. The robotic dog used in the trial was able to get back up and return to the starting line on its own in case of failure.