Science fiction is full of robotic systems that can fly. Whether they’re humanoid robots or robotic exoskeletons, all it seems to take to turn a walking robot into a flying robot is some thrusters attached to the hands and feet, and just like Iron Man, off you go. The reason that science fiction is full of this kind of thing is because it’s tremendously fun to think about this kind of thing, and you can imagine all kinds of compelling applications for it, even beyond the obligatory punching of aliens and more generalized anti-antagonist-ing.
In fact, some of these applications could be useful outside of science fiction, and researchers at the Italian Institute of Technology (IIT) are working on making it a reality, by developing a system that can control an iCub humanoid robot with four jet engines attached to it.
Getting an iCub to hover and fly by installing jet engines on its hands and feet sounds a bit far-fetched. But fundamentally, this is just another sort of multimodal locomotion, except it’s applied to humanoid robots rather than the hybrid platforms that we’re more accustomed to seeing. There are plenty of robots that combine flight with other kinds of locomotion to increase efficiency and versatility, so why not apply that to humanoids? Indeed, there’s a ton of work being done right now on aerial manipulation—using drones (sometimes drones with multi-DoF arms on them) to directly interact with the world around them. Well, you know what’s great at manipulation? Humanoids! So let’s just get them airborne already!
In a paper submitted to IEEE Robotics and Automation Letters (RA-L), Daniele Pucci, Silvio Traversaro, and Francesco Nori from IIT propose a “first step toward the development of a control framework for underactuated flying humanoid robots.” Their idea is that “there is a strong technological benefit in bringing humanoid and flying robots closer: a platform combining these two robot natures may have the capacities of flight, contact locomotion, and manipulation.”
We should emphasize here that when the authors call this a first step, it really is a first step. So far, what they’ve got is a basic control framework that can control a simulatediCub robot in a hover and gentle motion, relying on the thrust from simulated jet engines located where the robot’s hands and feet would be. The researchers are not considering aerodynamic effects, or any number of other very real and practical issues that an actual jet-powered flying iCub would need to be able to deal with. All of that comes later. For now, the goal is to use the simulation to verify that the control algorithm described in the paper has a reasonable level of robustness against modeling errors, which is necessary for its eventual implementation on a real robot.
The following video shows the simulated iCub hovering, as well as tracking a desired trajectory for its center of mass:
For more, we spoke with lead author Daniele Pucci:
IEEE Spectrum: Where did you get the idea for a flying iCub?
Daniele Pucci: I came up with the idea four years ago, when I first arrived at IIT. During my PhD, I developed a unified control approach for aerial vehicles, where the underlying flying robot was a single rigid body. So when I first saw the humanoid robot iCub, I wondered: How can I extend the theory in my PhD thesis, which is based on the single-rigid-body assumption, to the case of a multi-body articulated aircraft, as iCub? Everything started from this question.
In the paper, you say “there is a strong technological benefit in bringing humanoid and flying robots closer.” Can you talk more about that? What kinds of benefits would flying humanoid robots give us?
I believe that the benefits are many. First, there are technological benefits. Aerial humanoid robotics extends aerial manipulation to a more robust and energy efficient level. In fact, aerial manipulation is often exemplified by quadrotors equipped with a robotic arm. These robots can’t move around by means of contact forces with the environment, and they often struggle with flying in windy environments while manipulating an object, requiring precise position control for accomplishing manipulation tasks. So the extra hand of a flying humanoid robot could establish a contact point between the robot and the environment, thus making the robot position control simpler and more robust.
Another benefit is social. I truly believe that aerial humanoid robotics can be used as a test-bed for actuated flying exoskeletons for human beings. The recent successful story of Richard Browning shows the engineering feasibility of these futuristic actuated exoskeletons. However, the journey in front of us is still long, and we can use flying humanoid robots to boost this journey and avoid lots of tests on humans. We have invited Browning to IIT to discuss actuated flying exoskeletons, and he’s excited about our work on flying humanoid robots.
Finally, there are scientific benefits as well: In my humble opinion, controlling a flying humanoid robot leads to a number of theoretical and practical questions. For instance, a general control framework encompassing manipulation, contact-locomotion, and flight is still missing, and the role of the auxiliary (jet?) actuation during contact locomotion of humanoid robots is not clear. For instance, what is the walking speed at which it is more energetically convenient to turn the auxiliary actuation on? How do we deal with landing impacts for smooth transitions between flight and walking?
What are you working on next, and when do you think you will attach jet engines to an iCub robot to see if it will fly?
There are a few theoretical issues I would like to solve, like the convergence of the reference frame for the robot pelvis (or another general robot frame). Also, I’m investigating the aerodynamics on the humanoid robot, which is going to be fundamental if we want to equip the robot with wings for efficient horizontal flight. Then, considering aerodynamics in the control design is going to be one of the biggest theoretical questions.
But to make things work in practice, the problems are elsewhere. Characterising jet actuation, checking iCub’s torque control under that stress, and verifying that the inertial sensors will work properly under high level of vibrations are among the main challenges I envisage for the future. Right now, I’ve become the head of an IIT research team, which I call LOComotion Control Lab, so my time for research really shrunk. However, the main problem for defining a timeline is to find funding for aerial humanoid robotics. I really believe that once funding is available, in one and half years we may achieve some iCub hovering, but everything depends upon the starting time.
We thank Daniele for his answers. And somebody needs to fund this. Please. S.H.I.E.L.D., Stark Industries, Hammer Industries, whatever it takes. Let’s get this robot off the ground.