READY, JET, GO!: Flying humanoid robot takes off to enhance disaster response

Elyza Dimaapi 

It’s not a bird, it’s not a plane—it’s a flying robot. In a real-life leap straight out of science fiction, robots similar to Iron Man are no longer a fantasy. 

Researchers from the Italian Institute of Technology (IIT) made it possible through recently succeeding in their quest to build a robot equipped with a flying ability.

Known as iRonCub3, IIT’s jet-powered humanoid robot is the pioneer in the field. The prototype achieved overcoming the limitation primarily brought by changing environmental conditions.

Following its success, the researchers are looking towards the practical applications in disaster response, hazard inspection, and extreme environment exploration.

Published in Nature Communications Engineering, the IIT roboticist, in collaboration with the researchers of Polytechnic of Milan and Stanford University, shared that after two years of development, the prototype was able to hover 50 centimeters above the ground.

“This research is radically different from traditional humanoid robotics and forced us to make a substantial leap forward with respect to the state of the art,” explains Daniele Pucci, the Director of Artificial and Mechanical Intelligence at IIT.

iRonCub3 is a modified form of iCub3, a robot limited to walking, vision, and hearing. They reinforced the previous design with a titanium spine, added heat-resistant covers, and most importantly, installed four mini jet turbines. 

The engines can generate 1000 newtons of thrust, capable of lifting the 70-kilogram frame of the prototype.

On the other hand, the humanoid robot’s mechanical design is complete with arms where jet turbines are mounted, legs, and a torso with a jetpack containing two more turbines. 

This engine placement allows the robot to generate lift, to balance its weight, and to manage attitude control.

While in the development process, the roboticist team conducted a wind tunnel experiment and aerodynamics Computational Fluid Dynamics (CFD) simulations since the human-like body is not aerodynamic due to the long, uneven, and constantly shifting body parts.

“Our models include neural networks trained on simulated and experimental data and are integrated into the robot’s control architecture to guarantee stable flight,” explains Antonello Paolino, first author of the paper and a visiting researcher at Stanford University.

Moreover, the Deep Neural Network (DNN) developed by the researchers aims to correlate inputs from the dataset. It is a combination of the robot’s joint positions, the wind robot’s relative velocity direction, and other aerodynamic force components acting on the robot.

With the assistance of the neural network, the iRonCub3 can adjust its body mid-air, reacting to turbulence or wind the way a trained pilot might. 

As its limbs move, it rebalances automatically, using predictive AI to understand how each shift in posture affects the flow of air around its body.

Beyond its success, the researchers’ quest for a fully functional flying humanoid robot continues. 

With a common goal of enhancing disaster response, the prototype testing will continue to further enhance their innovation. 

The humanoid robot will take its next flight to the Genoa Airport, which will serve as a better equipped and safer ground for future experiments.