APRO : Teaching Drones to Go Unnoticed

One of the specialties of Benjamin Cotté, a faculty researcher at IMSIA—one of ENSTA’s 13 laboratories—is precisely aeroacoustics. In other words, he seeks to understand how airflow around a solid object produces sound. This is precisely the kind of challenge posed by drones, which is why he serves as the scientific lead for the APRO project (an acronym for “Aeroacoustics of Multi-PROpulsion Systems”), conducted under the auspices of the Defense Innovation Agency (AID) as part of the Interdisciplinary Center for Defense and Security Studies (CIEDS).

To precisely measure the sounds produced and eliminate any subjectivity in their perception, Benjamin Cotté and his team began by characterizing the noise emitted by drone propellers in their anechoic chamber. They tested the propellers individually, then in combination, at different rotational speeds.

Based on this, the researchers developed analytical models enabling a numerical simulation of these noises, which use aerodynamic calculations—taking into account the evolution of boundary layers (the interface zones between the blades and the air)—as input data to predict the radiated noise.

It was only after completing these first two steps that it became possible to consider concrete ways to reduce this noise.

“As for the propellers, we explored the possibility of modifying the surface roughness of the blades. Using 3D printing, we created blades with a very fine band of roughness—between 50 and 300 microns—at the point where the propeller’s profile is thickest. This effectively modifies the boundary layers. We demonstrated that this could be beneficial at certain rotational speeds, reducing noise by about 3 to 4 decibels. 
”

But drones never have just a single propeller. We must therefore account for possible interference between them, as well as interactions between the support arms and the propellers.

“We’re seeking to optimize the entire structure by taking into account as many geometric parameters as possible, such as the placement of the propellers relative to one another, rotor/stator interactions, and so on,” the researcher continues.

“Another interesting effect—if we take the case of a quadcopter—is that the rotational speeds of the four propellers are constantly changing to ensure the drone’s stability. We’ve therefore observed destructive or constructive interference effects between the propellers, which don’t always intersect at the same point. ”

However, the best way for an aerial drone to go unnoticed is still to stay away from people it might disturb. That’s the focus of researchers at U2IS, the lab on the Paris-Saclay campus of ENSTA’s Department of Information Sciences and Computer Science.

“For us, the challenge is to propose flight paths that reach the desired point but via a route that minimizes noise pollution,” summarizes Rémi Marsal, a postdoctoral researcher at U2IS.

An initial phase, led by Damien Hoareau, also a postdoctoral researcher at U2IS, involved classic flight path optimization.

Rémi Marsal then stepped in to propose using neural networks rather than a traditional optimization algorithm to simulate sound propagation. The network, trained on data from sound emitted by the drone, estimates sound propagation relative to a number of reference points.
 

In fact, in the cities of tomorrow, the future of drones may hinge less on their raw performance than on their ability to go unnoticed. Making a drone truly discreet without compromising its aerodynamic performance remains a challenge today. But the research conducted at ENSTA shows that there is room for improvement.

Spanning fluid mechanics, acoustics, materials science, and artificial intelligence, the quest for the silent drone draws on a wide range of expertise. This is a concrete example of the multidisciplinary approach developed at ENSTA to address the technological challenges of tomorrow.