Swarms of drones, a solution for regulating dense airspace

A team of researchers in Luxembourg is working on developing a unique system to protect airports from the disruption of drone attacks.

The project is being led by Grégoire Danoy, who heads the Parallel Computing & Optimisation (PCOG) group at the University of Luxembourg’s Interdisciplinary Centre for Security, Reliability and Trust (SnT). His team of researchers is exploring an approach inspired by living organisms: collaborative swarms of drones able to regulate low-altitude airspace and prevent dangerous or unauthorised situations, without relying on a single control centre.

A drone spotted too close to a runway can bring an entire airport to a halt. Flights delayed, traffic suspended, passengers stranded. These scenes, which have become familiar in Europe in recent years, illustrate a new reality: low-altitude airspace has become difficult to monitor and protect.

In these situations, the problem is not always the immediate danger posed by the aircraft, but the uncertainty it creates. Just a few dozen metres above the ground, the environment is dense and cluttered, leaving little margin for error. A single flying object can be enough to block an entire system.

At the same time, recent conflicts have highlighted other uses for drones, sometimes deployed in co-ordinated fashion and in large numbers. Whether civilian or military, these machines operate in a space that has nothing in common with conventional commercial aviation, and for which traditional regulatory tools are showing their limitations.

Faced with these new uses, how can we avoid risky situations without relying solely on centralised systems and without grounding airspace at the slightest doubt?

Low-altitude airspace difficult to control

Danoy has been working for over 15 years on these systems, where intelligence is not based on a single drone, but on the interactions between them all.

One of the main advantages of drone swarms lies precisely in their distributed organisation and the fact they do not rely on a single centralised system © Photo credit: SnT

At first glance, a swarm of drones may appear to be just a group of aircraft flying in a coordinated fashion. In reality, the difference is fundamental.

“A swarm is not just several drones flying together. It’s a distributed system in which collective behaviour emerges from local interactions between the drones,” Danoy explained.

The researcher often cites light drone shows to illustrate this distinction. Despite their spectacular appearance, they don’t operate on the same principle. “In these shows, the trajectories are entirely predefined. The drones follow a script and don’t react to their environment. So they are not swarms in the scientific sense,” he said.

In a real swarm, no single drone controls the others. Each device applies simple rules, based on what it perceives locally. These individual decisions give rise to coherent overall behaviour. A principle directly inspired by nature, from bird flights to ant colonies, which has long guided the PCOG group’s research.

The growing interest in these systems can be explained by the changing environment of low-altitude airspace close to the ground. This airspace is characterised by the proliferation of small flying objects, the presence of obstacles and a highly diverse range of uses. “Protecting this airspace is very different from regulating traditional air traffic,” Danoy said. “It is denser, more complex, with less margin for error”.

The disruptions observed around airports illustrate this point. “A small commercial drone can have disproportionate effects. Often, it’s not so much the potential damage as the uncertainty that leads to closing part of the airspace,” Danoy explained.

In their work, the SnT researchers are not directly addressing the detection of these objects. “That’s not our area of ​​expertise,” Danoy said.

Swarms of drones without a centralised decision-making centre

One of the main advantages of drone swarms lies precisely in their distributed organisation. “There is no central decision point,” Danoy said. “If a drone malfunctions or if communication breaks down, the swarm continues to operate, even if in a degraded state.” This resilience is a key property of these systems. It makes them particularly attractive for environments where communications are limited or unstable, such as certain rescue operations, remote areas or even space missions.

“In space, for example, repairing a breakdown can be very complicated, if not impossible. So these properties become essential,” Danoy said. Before reaching this point, the collective behaviours are extensively tested and validated. The work begins with theoretical modelling, followed by simplified simulations to quickly test ideas. This is followed by more realistic simulations, incorporating drone physics, communications and the environment.

These steps require significant computing resources. “We use the university’s high-performance computing infrastructure as well as national platforms. This is essential for generating and testing a large number of scenarios,” Danoy said.

Only after these digital tests do the experiments move to the real world. They take place exclusively indoors, in a controlled environment, with nano-drones weighing around 20 grams. “Today, we can fly about 15 of them, but in practice, our experiments often involve around 10 drones at the same time.”

The transition from simulation to reality remains one of the major challenges in the field. “This is particularly true when we use artificial intelligence to generate these collective behaviours,” Danoy said.

Military uses, cyber security and human responsibility

It’s hard to ignore the current geopolitical context, in which drones are widely used in armed conflicts. However, as Danoy points out, “our research into the protection of airspace began long before these events”. For him, the boundary between civilian and military use is not so much in the algorithms as in their deployment.

“Navigation, perception, coordination, autonomy: these are generic building blocks, necessary for applications such as infrastructure inspection, environmental monitoring or rescue missions. Their use then depends on regulatory frameworks, governance and societal choice,” he said.

Cybersecurity is also a concern. Although they are not specialists in the field, the researchers take into account the possibility of failures or interference. “For example, we are working on conceptual safeguards that strictly define what the autonomous system is allowed to do, so that it remains predictable and controllable, even in the event of a problem,” Danoy said.

The question of autonomy remains central. “For routine and well-defined tasks, such as mapping or environmental monitoring, autonomy can already go quite far and improve efficiency,” he said. In more sensitive contexts, however, this autonomy must be more closely regulated. “We can imagine adjustable autonomy, where the system operates independently for certain tasks, but requires human supervision for high-stakes decisions. Responsibility must remain with humans.”

At SnT, the goal is not to develop systems ready for field deployment. “Our role is to design algorithms, publish our results and advance research,” Danoy explained. “Operational applications will come later, driven by other stakeholders.” In a field where technological advances are rapidly evolving, drone swarms offer a unique testing ground at the intersection of fundamental research, security challenges and everyday civilian uses.

(This article was originally published by Virgule. Machine translated using AI, with editing and adaptation by John Monaghan.)