On July 11, Minister of Economic Affairs Dr. Nicole Hoffmeister-Kraut awarded prices to nine "AI Champions Baden-Württemberg 2024". One of the award winners is a researcher at the University of Stuttgart: Junior Professor Aamir Ahmad and his team are developing autonomous aerial robots for observing wild animals in the Wildcap project. The project thus makes an important contribution to species conservation.

[Picture: Wildcat / University of Stuttgart / Max Planck Institute for Intelligent Systems]

It started with robots playing soccer. "In my doctoral thesis, I dealt with interaction between robots using ball sports as an example," says Jun. Prof. Aamir Ahmad, Deputy Director of the Institute of Flight Mechanics and Flight Control (iFR) at the University of Stuttgart. In subsequent projects at the Max Planck Institute for Intelligent Systems (MPI-IS), Ahmad then sent teams of robots into the air: The civil engineer developed 3D motion capture systems in which several aerial robots equipped with cameras were used. "With the help of AI, we have trained the multi-robot system to recognize people and their postures in unstructured environments - i.e. in the wild. Several aerial robots track one person. They cooperate in planning their flight paths in order to get a better view of the person being tracked." The flying robots communicate with each other and, as a team, can create a much more meaningful overall picture from a large number of individual data points than a single robot.

Intelligent aerial robots in action for species conservation
Teams of autonomous aerial robots, for example, could help in the search for people who have had accidents, says Ahmad. A research trip to Kenya also revealed that the new technology has great potential for species conservation: "In Kenya, I spoke to a group of biologists from Princeton University. The researchers are developing strategies to protect endangered wild animals from extinction. As a basis for this, they need as much information as possible about the behavior of animals in the wild. However, the current methods of animal observation have various weaknesses. Camera traps, for example, cannot observe herds that cover large distances every day. GPS collars or tags attached to animals pose a risk, as the animals must be anesthetized for the procedure." The vision of developing intelligent aerial robots for animal observation was conceived in collaboration with the researchers in Kenya. The idea for the Wildcap project was born.

"Drones have also been used to observe animals in the past. But the approach of developing a team of intelligent aerial robots that recognize wild animals, follow them independently and evaluate their movements and behavior is unique so far," says Ahmad, who has held a junior professorship at the Institute of Flight Mechanics and Flight Control (iFR) at the University of Stuttgart since 2020 and heads the research group "Flight Robotics and Perception" there. At the same time, he also heads a research group at the MPI-IS.

Wildcap is financed by the Cyber Valley Research Fund. Robotics specialists from Stuttgart are working together with biologists and ecologists in Kenya and Hungary on this interdisciplinary project. Partners include the American Princeton University and the Hungarian Hortobágy National Park. "In Hungary, we are testing the use of aerial robots to monitor rare wild horses. In Kenya, we are tracking herds of zebras," says Ahmad. Observation from the air means fewer disturbance factors and therefore less stress for the animals than monitoring from the ground. The autonomous aerial robots can also cover much larger areas than camera traps, for example.

Robots are particularly impressive in a team
The Stuttgart team has developed various aerial robots for the project: Blimps, large balloon-like airships equipped with an intelligent control system and camera, are particularly energy-efficient and are suitable for long missions. Much smaller quadrocopters weighing just two kilograms are used to get closer to the animals. Commercially available drones were also retrofitted. These are used for short reconnaissance flights. "All our aerial robots work as a team and communicate with each other. With the help of formation flight strategies, we achieve the best possible visual coverage of the animals."

The team also relies on AI and machine learning methods. The systems are trained not only to recognize and track certain animal species, but also to classify their behavior. "For example, an aerial robot approaching a zebra determines whether the animal becomes nervous as it approaches. If this is the case, the robot could change its flight path independently." The systems could also draw important conclusions about an animal's activities from its posture: "Among other things, the robots recognize whether a zebra is grazing and collect data on how often and for how long an animal eats. In turn, this data can be used to draw important conclusions about the zebra's health."

​https://www.uni-stuttgart.de/en/university/news/all/Zebras-wild-horses-and-teams-of-flying-robots/