AI-driven autonomous robots achieved unprecedented insights into the heart of the bee colony through 24/7 close-ups of the queen and her routine.

Scientists at the Artificial Intelligence Centre FEE CTU in Prague and researchers from the University of Graz, Durham University, and the Middle-East Technical University in Ankara have developed a game-changing technology for a novel type of digitized behavioral research. Their autonomous robots, operating in a honeybee research facility located on the campus of the University of Graz, generate 1.4 terabytes of data streams every day. These data contain high-resolution real-time images of the queen bee and her interactions with worker bees for further analysis by partner institutes of the EU.

In the current article Autonomous tracking of honeybee behaviors over long-term periods with cooperating robots, published in the renowned journal Science Robotics, the doctoral student Jiří Ulrich and his co-authors present the multitude of relevant data that the robotic system collects. In addition to the queen's movement patterns and egg-laying behavior, it can count the population, measure breeding success, and continuously monitor the brood, pollen & honey on the combs.

Typical image captured with the robotic system, as presented in the Science Robotics article. In the center is the queen bee, marked with a special tag. Around her are the "court bees" that tend to and feed her.

Due to the high complexity of the bee colony, the observation of the queen requires assessing the behavioral context, such as the overall hive status, by recording the bee population and the development of the comb contents in order to allow a fundamental understanding of the social self-regulation of the bee state.

For the first time, the new robotic system achieved highly accurate and continuous long-term collection of a wide range of relevant data in the beehive simultaneously. The researchers demonstrate the versatility of the research facility by exemplifying 23 so-called "Key Behavioral Metrics," or essential quantifiable behavioral patterns, emphasizing that more analyses will be added throughout the RoboRoyale project.

Special Issue on Robots for Scientific Discovery.

The article reports on some astounding initial findings. The researchers found out that the queen bee observed in this study covers a distance of about 1.5 kilometers in a month (on two honeycombs approx. 42x33cm in size), that she prefers to rest at specific areas for up to 1.5 hours, and even in the waning bee season of October, lays an average of 187 eggs per day.

The robotic system was built around a fully functioning bee colony, allowing the bees to carry out their regular foraging flights to the surrounding plants. When the queen walks, the robots move in concert with her in focus, even when she walks to the other side of the comb. When the queen rests, one of the cameras keeps monitoring the bees that groom her while the other camera performs a sweep of the comb to count the worker bees as well as the brood, pollen, and honey cells of the comb.

System dashboard showing two sides of the observation hive. The current trajectory of the queen is in red, the past trajectory in yellow. The positions of the loaded eggs are marked with purple dots. At the bottom is information about the activity of the queen and her court.

The recordings use infrared light to avoid influencing the bees' behavior. The HD-quality recordings are subsequently analyzed using advanced computer vision methods. The study comprised more than 100 million individual images, a practically unmanageable volume for human scientists to analyze manually.

With this new system, the international team led by Tomáš Krajník showed that intelligent robots are a game changer in behavioral research and ecology as they enable unprecedented insights and knowledge into these complex systems.