Nature as inspiration

How drones are learning to swarm at the University of Augsburg


The animal kingdom has often inspired technical innovation. Swarms of birds, fish, and insects are fascinating natural spectacles, which have for a long time captured the interest of researchers. While the reasons behind natural swarming behaviour are more well understood as a consequence of years of research, the possibilities for the use of swarms in technical systems is less well-known. At the Institute for Software & Systems Engineering at the University of Augsburg (ISSE), Prof. Dr Wolfgang Reif and his team are researching how swarming behaviour can be used with flying drones.

© University of Augsburg

At first, a black screen with many coloured dots appears on Dr Oliver Kosak’s computer monitor, an expert in self-organising systems at the ISSE and topic coordinator at the AI Production Network. When he starts the simulation, the dots start buzzing around. What looks chaotic at first quickly becomes ordered. The dots of each colour sort themselves around a centre so that they come to represent the shape of the Olympic rings. This simulation became a reality during the 2021 Olympic Games in Tokyo, with drones flying in the pattern of the Olympic rings against the backdrop of the night sky. “You can see that a huge amount of planning goes into this and other light shows, and how much work is necessary to design all the flight plans for each individual drone in advance,” explains Kosak.

Rules of behaviour for swarming drones

To reduce the amount of planning required for such projects, the Chair of Software Engineering is researching self-organising systems through a DFG funded project. “A paradigm shift has occurred so that the drones no longer have to be controlled individually but can communicate with each other to achieve their goal without the need for help from outside,” Reif explains. Animal swarms serve as a model for such organisation, which have been studied in biology for a long time. Knowledge from this research has provided insights for understanding the processes of physics and system dynamics, which can also be used in computer science. The team of researchers at Augsburg have been able to apply principles gained from biology to the simulation of drone swarms. Individuals always want to stay in the swarm, so they orient themselves to their neighbours and try to fly like them, having the same speed, but not colliding,” explains Reif.

From simulation to reality

Individual drones have been used in everyday life for some time now. They deliver packages and inspect industrial facilities such as wind turbines and solar parks. Aside from light shows, however, swarms of drones have not been utilised all that much. “This is mainly because research is still needed on the control mechanisms,” says Kosak. Together with Dr Constantin Wanninger, head of flight robotics at the ISSE, the practical part of drone research is being conducted in a flight arena, where techniques tested in simulations are then transferred to real drones. “With a lot of work and some adjustments, you can see that it can work with real systems, even if it is not as seamless as the simulations,” says Kosak. The first field tests have already taken place, in which the swarming drones held a fibre optic cable several metres above the ground in order to measure the ground temperature. “This provides a very mobile measuring instrument that can cover a large area, which can even be used at altitude.”

Using swarming drones after catastrophes

The use of swarming drones following catastrophes is also being researched. These could be used following chemical accidents, for example, in order to determine the movement and direction of toxic fumes more accurately. The closest stationary weather stations, which can accurately determine the direction of the wind, are often located upwards of several kilometres away from the location of accidents, which means the effects of the weather on the spread of fumes is often unknown. According to Kosak, swarms of drones could be used to help make decisions based on their identification of the position with the highest concentration of gases, thereby determining whether the spread of the fumes could pose a critical risk to the population.

Our researchers

Assistant Professor
Institute for Software & Systems Engineering
Institute for Software & Systems Engineering