In Delft, researchers at the Delft University of Technology have opened the nation’s first dedicated facility for studying autonomous miniature drone swarms. Known as the Swarming Lab, the initiative seeks to develop a coordinated fleet of 100 self-flying drones capable of performing continuous operations without human intervention. These drones, each weighing roughly the same as a golf ball or an egg, are designed to land autonomously on charging pods, recharge, and resume flight in a closed operational loop.
Guido de Croon, director at the Swarming Lab, explained the core ambition: “We are working not only to get these robots to be aware of one another but also work together to complete complex tasks.” Among the envisioned applications is industrial safety—swarms equipped with gas sensors could patrol factory environments until one detects a leak. That drone would then follow the gas trail while summoning others to assist, enabling rapid localization of hazards. Similar principles could be applied to detecting forest fires or conducting large-scale search-and-rescue missions.
The team draws heavily on biological models, studying how bees, ants, and flocks of birds coordinate. “Drone swarm technology is the idea that when we look at nature and you see many of these animals like ants, that individually are perhaps not so smart, but together they do… things that they could definitely not do by themselves,” said De Croon. He added, “We want to instil the same capabilities also in robots.”
Bird flocking behavior offers a particularly useful template. Birds monitor their nearest neighbors, maintaining distance to avoid collisions while staying close enough to remain part of the group. They align their movements, creating fluid patterns that serve purposes such as predator evasion. “By following such simple rules you get these beautiful patterns that are very useful for the birds, also against predators,” De Croon noted. Translating these simple behavioral rules into robotic algorithms is central to the lab’s approach.
Yet the complexity of swarm dynamics presents significant challenges. “Swarms are complex systems,” De Croon said during a demonstration at the TU Delft Science Centre. “A single robot can do simple things within a swarm. It is actually quite difficult to predict, however, with these simple rules how a whole swarm will behave.” This unpredictability requires careful modeling and iterative testing.
Hardware constraints add another layer of difficulty. The small form factor limits onboard sensor arrays and computational resources. At present, the Swarming Lab’s drones depend on an externally mounted camera system to track their positions relative to the swarm. However, progress is being made toward enabling drones to sense one another autonomously, without external infrastructure. Such capabilities would greatly enhance operational flexibility, particularly in environments where external tracking is impractical.
Comparable work has been demonstrated elsewhere. In 2022, researchers at Zhejiang University in China successfully flew 10 autonomous drones through a dense bamboo forest, showcasing the feasibility of navigating complex natural terrain without centralized control. This milestone underscores the global momentum in swarm robotics research.
The Swarming Lab collaborates closely with Emergent, a start-up founded by former TU Delft students. Together, they currently operate a fleet of about 40 small drones. Lennart Bult, co-founder at Emergent, outlined their trajectory: “The aim is eventually to put a swarm of around 100 drones in the air in the next five years.”
For De Croon, the ultimate goal extends beyond technical benchmarks. “It would be really great if we actually get a bit closer to the astonishing intelligence of tiny creatures like honeybees,” he said. By merging insights from biology with advances in robotics, the Swarming Lab aims to unlock new capabilities in autonomous aerial systems, potentially transforming how drones are deployed in industrial, environmental, and humanitarian contexts.