With a few exceptions, quadrotors are the go-to aerial robot when you need something small, fast, and maneuverable. This is because quadrotors are relatively cheap and easy to fly, and not because they’re the best aerial platform. In fact, you may have noticed that there aren’t a lot of rotary fliers in the animal kingdom—this is because (among other reasons) flapping wings offer high efficiency and incredible maneuverability as long as you’re able to manufacture and control them.
Those last two things are what make wings tricky for robots, which is why we don’t see nearly as many useful robot birds as we do useful quadrotors. Alireza Ramezani, Soon-Jo Chung, and Seth Hutchinson from the University of Illinois at Urbana-Champaign and Caltech have decided that making robot birds is just not tricky enough, so they’re working on something even better and even more complex: a robotic bat.
We’ve been keeping track of Bat Bot (B2) for the past few years; the researchers presented a paper on the robot at ICRA 2016 last year in Stockholm. Now B2 has made it onto the cover of the current issue of Science Robotics, and that’s cool enough for us to publish an update. Here’s the video:
Bat wings are fundamentally different from bird wings, and it’s not just because birds have feathers and bats don’t. Generally, when roboticists design bird-inspired or insect-inspired robots, they use rigid approximations of the wings, or perhaps a few different rigid parts flexibly interconnected. Bat wings don’t work like this at all: The underlying structure of a bat’s wing is made up of “a metamorphic musculoskeletal system that has more than 40 degrees of freedom” and includes bones that actively deform during every wing beat. The wing surface itself is an “anisotropic wing membrane skin with adjustable stiffness.” This level of complexity is what gives bats their unrivaled level of agility, according to the researchers, but it also makes bats wicked hard to turn into robots.
Rather than trying to replicate every single degree of freedom (DoF) of a real bat (which probably would have resulted in a robot that would have been too heavy and complicated to fly), the researchers cut down the number of DoFs to five (the shoulder movement, elbow movement, wrist bending, and side-to-side movement of the legs and tail), which still allows the robot to theoretically replicate over 57 percent of bat flight kinematics. B2 is about the same size as an Egyptian fruit bat, with a wingspan of 47 centimeters and a weight of 93 grams.
B2 uses onboard custom electronics to be able to fly autonomously. The navigation and control algorithms run on the main control board in real time, and a separate data acquisition unit helps to process sensor data and command the robot’s actuators. The sensors include an inertial measurement unit (IMU) and five magnetic encoders located at the elbows, hips, and flapping joints. Image: Science Robotics
Part of the reason that so many DoFs are necessary in B2 is that the wing surface morphs (stretches, flexes, and twists) with every movement of the underlying structure, which can have drastic aerodynamic consequences. The wing itself is made of a flexible silicone membrane which is just 56 micrometers thick. Controlling B2’s DoF such that the wing surface morphs into an aerodynamic configuration that results in the robot doing what you want it to do is an enormous challenge, which the researchers solved using closed-loop feedback. They’ve managed to get B2 to autonomously stabilize, cruise (fly straight while flapping its wings at up to 10 hertz), turn in a bank, and dive sharply, and this technique “can potentially help reconstruct the adaptive properties of bat flight for other maneuvers.”
Bats can do some absolutely amazing things: Besides being able to perch upside down (which is tricky if your initial condition is rightside up and flying), they can actually catch insects in their wings and carry them back home. The researchers mentioned both of these capabilities, saying that they’re specifically working on the upside-down perching thing, but our guess is that catching insects (or anything else) in midair is probably not going to happen soon, especially if the robot is intended to keep flying after it happens. But while B2 may not be able to replicate everything that a real bat can do (yet), it’s already helping us to understand how real bats work: You’ll have a lot of trouble trying to convince a real bat to fly the same path 10 times in a row to see how it moves its wings to maneuver, but a robot will quite happily do all of the experiments you could ever want.
Eventually, B2 (or bat robots like it) could be used in situations where flight in close proximity to obstacles is necessary, since wings are much more forgiving to collisions that rotors are. The researchers are specifically interested in construction site inspection, although they also talked about such robotic research project staples as disaster relief and surveillance. It’s hard to guess when we might see a commercial robotic bat, but it’s certainly going to require more developments in small and lightweight batteries, actuators, and computers, and hopefully the researchers will keep on presenting updates at ICRA and IROS so that we can see how Bat Bot evolves.