Designing drones that can fly in air ducts

Inria Nancy-Grand Est

1. Air ducts: a contemporary issue for safety inspection

Air ducts are everywhere in modern buildings and underground networks, but are challenging to access for inspection. Their narrow dimensions and inability to support human weight prevent essential interventions to maintain air quality, heating, and air conditioning.

Small quadrotor drones offer a potential solution for exploring these air ducts because they can navigate both horizontal and vertical sections. However, they create airflows that recirculate inside the duct and destabilize the drone, creating important turbulences in an environment that has little space for error.

1. Mapping the aerodynamic forces in a circular air duct

The team first investigated how the air circulates depending on the position of the drone in the air duct. To do so, they used a robotic arm and a force/torque sensor to measure the forces at hundreds of positions. This first “map” of forces shows a complex aerodynamic pattern and makes it possible to identify the “unsafe” parts of circular ducts, where the air recirculations push the drone towards the walls, and a safer position, where the recirculation forces cancel out.

1. Flying and hovering small quadrotors in small air ducts

To maintain its position at the recommended point, the drone needs to know its current position in an environment which is typically dark and without any visual cues. The team combined small lasers and artificial intelligence (a neural network trained on motion capture data) to allow the small drone to stay in the position with the smallest turbulences, hence flying in a safer and more stable way.

1. A step toward many applications

The results of this research open new and promising application domains for drones in industrial inspection and public safety. The next step is to develop a more application-oriented prototype with useful payloads, like cameras, thermal cameras, or gas sensors.

1. A large-scale collaborative project

This study was led by Inria Senior Researcher Jean-Baptiste Mouret and PhD student Thomas Martin, both from the project-team HUCEBOT, a joint venture between CNRS, Inria and Université de Lorraine, within the Centre Inria at the University of Lorraine and the Lorraine Research Laboratory in Computer Science and its Applications (CNRS/University of Lorraine). It is a collaboration with the Institut des sciences du Mouvement - Etienne-Jules Marey (Aix-Marseille Université/CNRS) and the Laboratoire de Conception, Fabrication, Commande (Université de Lorraine). This work reflects the strength of the partnerships between the French institutions that are Aix-Marseille Université, CNRS, Université de Lorraine, and Inria.

1. References

Scientific article: https://www.nature.com/articles/s44182-025-00032-5

Citation: Thomas Martin, Adrien Guénard, Vladislav Tempez, Lucien Renaud, Thibaut Raharijaona, Franck Ruffier & Jean-Baptiste Mouret. Flying in air ducts. npj Robotics 3, 16 (2025). https://doi.org/10.1038/s44182-025-00032-5

1. Media (videos, photos)

Video: https://www.youtube.com/watch?v=7je2hUnGwms

 

---

Journal

npj Robotics

DOI

10.1038/s44182-025-00032-5 

Method of Research

Experimental study

Subject of Research

Not applicable

Article Title

Flying in air ducts

Next generation of autonomous drones will harness wind like an albatross

University of Cincinnati researchers are developing drones that can optimize wind in real time

University of Cincinnati

LinkedInWeChatBlueskyMessageWhatsAppEmail

 

image: 

A black-footed albatross soars over the Pacific Ocean. Engineers are working to mimic the amazing ability of albatrosses to harness the wind for the next generation of autonomous drones.

view more 

Credit: Michael Miller

How does one of the biggest birds in the world spend so much time in the air?

Albatrosses have 11-foot wingspans that carry them across oceans. But it’s how they use these wings that makes them world-class flyers, according to a University of Cincinnati aerospace engineering professor.

UC Assistant Professor Sameh Eisa and his research partners hope to harness their amazing abilities for the next generation of drones.

He received a $700,000 grant from the Defense Advanced Research Projects Agency, part of the U.S. Department of Defense, to develop innovations in unmanned aerial vehicles using animal-inspired engineering called biomimicry.

The project is based on his recent breakthroughs in developing model-free, real-time flight controls to harness the natural abilities of the albatross. 

Albatrosses use a principle called dynamic soaring to master the wind for both distance and time in the air. Eisa and his team developed a first-of-its-kind approach to dynamic soaring they call “a natural extremum-seeking system” after the way the birds (and their drone-mimic) identify the minimum and maximum pitch, yaw and roll and air speeds needed for optimal efficiency.

The birds tack into the wind like a sailboat to gain lift and altitude, finding faster air currents as they climb. When they eventually lose the forward momentum needed to stay in the air, they turn, harnessing the kinetic energy of gravity and wind that propels them forward. At the bottom of this glide path, sometimes mere inches off the water, they turn back into the wind and do it again — all without wasting a single wingbeat.

“They use it skillfully,” Eisa said. “GPS trackers show these birds can fly hundreds of miles a week. By the time they die, they’ve flown 20 times the distance between the Earth and the moon.”

But Eisa said there’s more to the bird’s energy-efficient flight than its enormous wings.

“Albatrosses literally have a nose for wind,” Eisa said.

The birds are able to gauge wind speed and direction through their sensitive nostrils, allowing them to make fine flight adjustments to maximize each leg of upward and downward flight path.

Eisa’s analyses show that energy from the wind balances the energy traditionally lost in flight. Meanwhile, the total energy of each dynamic soaring cycle is near constant. An applied mathematician, Eisa put the birds’ abilities to the test in simulations and found that computers could do no better at charting the optimal course in real time.

“They are solving an optimization problem that is unbelievably complicated,” Eisa said.

For a drone to fly like an albatross and achieve autonomous soaring, it will have to measure both changing wind speeds and direction to calculate the best angle of attack and rolling action to adjust flight controls in real time, he said.

“If we can get closer to how the albatross does it, we can be more efficient,” he said.

Eisa and his students are collaborating with researchers in industry, weather experts and the Massachusetts Institute of Technology on the project called Albatross.

Traditionally, wind is the enemy of drones, Eisa said. But their project is trying to turn this obstacle into an advantage.

Using Eisa’s recent characterization for dynamic soaring as a natural extremum-seeking system, new flight control designs will be developed to mimic dynamic soaring in real time. Researchers will test, validate and implement these designs and methods in experiments by UC’s DARPA industrial team members to demonstrate how much energy dynamic soaring saves compared to normal flight.

That’s one reason biomimicry has been such an important tool for aerospace engineers, he said.

“Nature has been optimizing flight for millions of years of evolution,” Eisa said. “So to take this gift from nature and make it available to humanity is engineering at its best.”

Aerospace engineering students are developing the next generation of unmanned aerial vehicles in UC's College of Engineering and Applied Science.

Credit

Andrew Higley

The wandering albatross has the largest wingspan of any living bird. They make acrobatic turns in their energy-efficient flight. Aerospace engineers are working to mimic their amazing ability to harness the wind for the next generation of autonomous drones.

Birds like this black-browed albatross employ dynamic soaring to save energy in flight. Aerospace engineers at the University of Cincinnati are designing autonomous drones that can identify wind speed and direction and make flight adjustments in real time to save energy.

Credit

Michael Miller

How brain-inspired analog systems could make drones more efficient

Electrical and computer engineers want to mimic the brain’s visual system to create AI tools for guiding autonomous systems.

University of Rochester

LinkedInWeChatBlueskyMessageWhatsAppEmail

The artificial intelligence systems that guide drones and self-driving cars rely on neural networks—trainable computing systems inspired by the human brain. But the digital computers they run on were initially designed for general-purpose computing tasks ranging from word processing to scientific calculations and have ultra-high reliability at the expense of high-power consumption.

To explore novel computer systems that are energy efficient particularly for machine learning, engineers at the University of Rochester are developing new analog hardware, with the possible application toward more efficient drones. Rochester engineers are attempting to do so by abandoning conventional state-of-the-art neural networks developed on digital hardware for computer vision. Instead, they’re turning to predictive coding networks, which are based on neuroscience theories that the brain has a mental model of the environment and constantly updates it based on feedback from the eyes.

“Research by neuroscientists has shown that the workhorse of developing neural networks—this mechanism called back propagation—is biologically implausible and our brains’ perception systems don’t work that way,” says Michael Huang, a professor of electrical and computer engineering, of computer science, and of data science and artificial intelligence at Rochester. “To solve the problem, we asked how our brains do it. The prevailing theory is predictive coding, which involves a hierarchical process of prediction and correction—think paraphrasing what you heard, telling it to the speaker, and using their feedback to refine your understanding.”

Huang notes that the University of Rochester has a rich history in computer vision research and that the late computer science professor Dana Ballard was an author on one of the earliest, most influential papers on predicative coding networks.

The Rochester-led team includes Huang and electrical and computer engineering professors Hui Wu and Tong Geng, their students, as well as two research groups from Rice University and UCLA. The team will receive up to $7.2 million from the Defense Advanced Research Projects Agency (DARPA) over the next 54 months to develop biologically inspired predictive coding networks for digital image recognition built on analog circuits. While the initial prototype will look at classifying static images, if they can get the analog system to approach the performance of existing digital approaches, they believe it can be translated to more complex perception tasks needed by self-driving cars and autonomous drones.

And while the approach is novel, the system will not use any experimental devices but will instead be manufactured using existing technologies like the complementary metal oxide semiconductor (CMOS).