Collective sensing and motion of bats

Imagine being inside a cave. There is no light. You are surrounded by absolute darkness. This is Thejasvi Beleyur’s research setting: the Orlova Chuka cave in Bulgaria. Beleyur is interested in how animals move and sense each other when they are in groups and have limited information about their surroundings. He studies groups of echolocating bats that jam each other due to their loud calls, but still manage to listen to faint echoes to detect each other.

An important aspect of Beleyur’s research is that he studies bats in their natural environment. He uses acoustic as well as video-tracking systems and matches the bats’ behaviour to their position in space by aligning the tracks with LiDAR scans of the cave. LIDAR thermal aligning has already been done in ‘domestic’ settings, such as offices, buildings, and cityscapes. But alignment in the wild, such as in caves, has not been attempted. “The complex shapes and thermal gradients in natural scenes required some extra processing, and computer scientist Bastian Goldlücke and I learned a lot about how well the existing algorithms work,” says Beleyur.

Did the effort pay off? Beleyur confirms that it did: “Many species will not show the same collective behaviours under lab conditions. The so-called Ushichka dataset – an ode to the place where the data were collected – overcomes both these limitations, as it collects multi-sensor data of up to 30 bats flying around happily within natural conditions.” He continues: “Echolocating bats form extremely large gatherings and show some of the most impressive examples of collective behaviour. I argue that they have only been studied in one or two modalities at most in terms of the data collected, such as audio or video. Even lab experiments can only go so far regarding the number of bats you can put in a room.”

“The so-called ‘beamshapes’ package could be used to study acoustic communication and model sound radiation in birds, humans, meerkats, etc,” says Beleyur. The methods developed in this project can therefore be used for other model systems. It is also published as an open source implementation. In the future, Beleyur plans to perform collective behaviour modelling based on this data.

What are some interesting outcomes of this experiment? “The challenge when groups of bats move together is that they effectively ‘blind’ each other,” says Beleyur. “Each bat emits really loud calls that prevent other bats from hearing echoes. Modelling shows that the bats are probably flying in a ‘hyper-stroboscopic’ reality, where they detect each other only every now and then because of all the mutual deafening from the loud calls. Not much is known about how bats actually show collective behaviour, despite the fact that each individual experiences this ‘hyper-stroboscopy’ when in groups.”