Squirrels and the US Army are an unlikely combination at best. Research at the University of California, Berkeley is trying to provide an advantageous link between them, at least as far as the Army is concerned. The squirrels could probably care less!
It reminds me of the old “Golden Fleece Award” that Senator Proxmire invented in the 1970’s.
From 1975 to 1988 Senator William Proxmire issued 168 “Golden Fleece Awards” for what he considered spurious uses of public funds. It was a Washington institution, which he based on the 15th century chivalric award of the same name, and the play-on-words between the two meanings of “fleece”.
Specifically, researchers at Berkley are studying fox squirrels in an effort to understand why they’re so damn good at making treacherous leaps to reach delicious peanut treats. In fact, squirrels are so good at “reading” obstacles with their own physical limitations in mind that they almost never fall.
Understanding the split-second decisions they make as they jump from branch to branch will help scientists develop more agile robots, at least according to the U.S. Army.
Locomotion is a complicated sub-field of robotics for a reason: Machines are pretty good at falling over and getting stuck. So, as the thinking goes, a bio-inspired approach, using the ubiquitous backyard squirrel as a test case, could help military, and search-and-rescue, robots negotiate even the most uneven terrain.
While the U.S. military has fielded a broad array of flying drones over the past two decades, from the original MQ-1 Predator to the passenger-jet-sized MQ-4 Global Hawk, the development of ground-based drones has lagged. Unlike flying drones, which rarely need to tackle obstacles or tricky situations, ground drones roam in a much more complex environment. Ground drones can face an array of obstacles, some navigable and some not. Detecting whether it can climb a pile of rocks or jump across a trench, for example, are the kinds of tricky questions a robotic brain must be able to answer quickly in the field.
For decades, roboticists have studied other living creatures, such as the gecko and the cockroach, in their quest to build agile robots that can conquer challenging situations. But now the focus has shifted to an even dicier problem: figuring out how robots can learn to make split-second locomotion decisions based on their own mechanical limitations.
Enter the squirrels. Squirrels and the US Army.
In a eucalyptus grove on the UC Berkeley campus, as part of a project funded by the U.S. Army, the National Science Foundation, and the National Institutes of Health, researchers have been studying the wild rodents, manipulating them into making jumps for treats.
“If we want robotic platforms to go wherever the fight takes them, these platforms will require fast, creative decision making,” said one of the researcher team.
Squirrels possess the incredible ability to judge an obstacle with their own agility in mind. They can study a problem, such as reaching a peanut suspended above the ground, and determine if a course of action, like a leap from a log to the treat, will be successful. Squirrels aren’t always perfect in judging a jump, but they never experience catastrophic failure. While they might pause to assess and even make a few false starts, the squirrels never actually fall while making the jump. In the worst-case scenarios, they can also use their claws to firmly latch onto trees.
Maybe they should try training squirrels instead of trying to get robots, or soldiers, to act like them. The next “Golden Fleece” awaits squirrels and the US Army.
