Leveraging big data and deep learning to uncover the connection between biosonar sensing and flapping flight in bats

Understanding the coupling of biosonar sensing and flapping flight in bats poses a challenge, because control of the many degrees of freedom in the flight apparatus of bats is not readily reconciled with the low-dimensional nature of the biosonar inputs. This is particularly true for bat groups that are capable of dexterous maneuvers in dense vegetation and rely on biosonar as their primary source of sensory information about their environments. To uncover how bats with sophisticated biosonar sensing and agile flight are able to control their aerial maneuvers based on the biosonar echoes, a laboratory setup with synchronized arrays of high-speed video cameras and ultrasonic microphones has been established to collected large volumes of data. The hardware setup is complemented by deep-learning methods to process the raw data from the arrays and discover informative patterns. The effectiveness of these methods opens the possibility to leverage the behavioral variability across individual and species for the discovery of common patterns and adaptations. To ensure that insight into biosonar and flapping flight of bats can be exploited for advances in bioinspired technology, biomimetic sonar and flapping-flight robots are being recorded under the same conditions as the bats to allow for side-by-side quantitative comparisons.