RHex-type legged robots are great at getting around. Like, really, really, really great. They can walk and run on land, and a RHex-based hexapod called Aqua can swim in water as well, with just a simple change of legs from something rigid for walking to something flexible for swimming. Technically, this makes RHex amphibious, but in practice, it's more like the robot is amphibious if you've got a human around to swap its legs out. The problem is that it's impossible to make legs that are flexible enough for efficient swimming and simultaneously rigid enough for efficient walking. And when we say "impossible," we mean "impossible until someone figured out how to do it," which happened at IROS last month.
These new amphibious legs from McGill University roboticists—the creators of the Aqua robot—are called "Ninja legs," because the researchers figure that "the design resembles a spinning ninja star." I'm not sure I entirely get that, but it's a cool name anyway. Essentially, Ninja legs are flippers that can flex up to 120 degrees, each contained inside a carbon fiber and spring steel frame that are mounted on Aqua's rotary leg joints . On land, the frame works just fine as a leg, while in the water, the frame allows the flipper inside it to move freely, resulting in efficient swimming.
As is usually the case when you take two very different things and combine them into one thing, the Ninja legs involve some compromises. The most obvious disadvantage is perhaps complexity, and the efficiency for both walking and swimming decreases relative to legs that are designed and optimized specifically to perform one of those tasks. This decrease is not nearly as bad as might be expected, though. For walking gaits, the ninja legs require perhaps 15 to 20 percent more power to rotate at a given frequency, but they can also achieve higher stable speeds due to higher inherent compliance. And for swimming, the Ninja legs hit a sweet spot at a 2.5 hertz oscillation frequency and 50 degree oscillation amplitude where they produce nearly as much thrust as flippers can (35 N as opposed to 40 N).
There is a lot more that can be done as far as figuring out whether the Ninja legs might do better with a completely different gait than a more traditional RHex configuration, so it's possible that they could get even more efficient. And there's also plenty of space for hardware optimization. The future work that we're most looking forward to, however, is some more field testing to see just what hexapods with Ninja legs are capable of.
"Ninja Legs: Amphibious One Degree of Freedom Robotic Legs," by Bir Bikram Dey, Sandeep Manjanna, and Gregory Dudek from the Centre for Intelligent Machines at McGill University, was presented last month at IROS 2013 in Tokyo, Japan.