Fascination With Robotics How This Humanoid Robot Learns to Fly

From Sandro Kipar | Translated by AI 2 min Reading Time

Related Vendors

In our "Fascination with Robotics" section, we present extraordinary robots to you once a month. Today: A humanoid robot learns to fly.

Stabilizing a humanoid robot in flight is a major challenge because the center of gravity can quickly change. However, with the help of AI, the scientists were able to overcome initial hurdles.(Image: IIT-Istituto Italiano di Tecnologia)
Stabilizing a humanoid robot in flight is a major challenge because the center of gravity can quickly change. However, with the help of AI, the scientists were able to overcome initial hurdles.
(Image: IIT-Istituto Italiano di Tecnologia)

With the iRonCub3, humanoid robotics is literally taking off into new dimensions. Developed by the Advanced Mechatronics and Interaction Laboratory (AMI-Lab) of the Istituto Italiano di Tecnologia (IIT) in Genoa, the latest prototype combines state-of-the-art humanoid mechanics with jet turbine propulsion and intelligent flight control systems. The result is a robot that can not only walk and grasp but also actively fly. The researchers first reported their findings in Nature Communications Engineering.

Weighing around 154 pounds, the iRonCub3 is based on the third generation of the iCub robot but has been thoroughly revised. A total of four microturbines are integrated into the jetpack and in both forearms, collectively generating over 1000 Newtons of thrust. The design has been entirely adapted to the thermal and mechanical requirements of a flying robot: high-temperature-resistant materials, a titanium spine for load distribution, and specially designed casings protect the internal systems from the turbines' exhaust gases, which exceed 1470°F.

One of the biggest challenges, according to the researchers, was developing an intelligent flight control system. The team relies on a combination of model predictive control (MPC), optimization algorithms, and sensor-based state estimation using inertial measurement data and real-time camera technology. The system is complemented by a deep-learning model capable of capturing and compensating for aerodynamic effects in real-time, using wind tunnel data and Computational Fluid Dynamics (CFD) simulations. This is made possible through a co-design approach, where form, jet positions, and control architecture are optimized together.

"This research radically differs from traditional humanoid robotics and forced us to make a significant leap beyond the current state of the art," says the study leader Daniele Pucci. "Aerodynamics must be assessed in real-time while the control systems must master both slow joint actuators and fast jet turbines. Testing these robots is as fascinating as it is dangerous—and improvisation is out of the question."

Controlled Hovering at 20 Inches Height

The first controlled test flight of the iRonCub3 took place on the roof of the CRIS-Lab: the robot flew at a height of about 50 centimeters with visible stabilization across all axes. Further flight tests in professionally secured environments, including Genoa Airport, are already planned. In parallel, wind tunnel tests are being conducted in cooperation with the Polytechnic University of Milan to further optimize aerodynamic behavior under different load profiles.

For the industry, this opens up completely new perspectives: a humanoid flying robot like iRonCub3 could be used in the future for inspecting hard-to-reach industrial facilities, during emergency operations in dangerous environments, or when accessing destroyed infrastructures. Thanks to its humanoid form, not only does the movement logic remain familiar, but the ability to grasp tools or manipulate controls is also retained—a capability that pure drone solutions cannot currently offer.

Subscribe to the newsletter now

Don't Miss out on Our Best Content

By clicking on „Subscribe to Newsletter“ I agree to the processing and use of my data according to the consent form (please expand for details) and accept the Terms of Use. For more information, please see our Privacy Policy. The consent declaration relates, among other things, to the sending of editorial newsletters by email and to data matching for marketing purposes with selected advertising partners (e.g., LinkedIn, Google, Meta)

Unfold for details of your consent