Good vibrations: vortex encoding of the seal whisker array

Behavioral experiments of seals have demonstrated the animal’s ability to learn various new tasks based on whisker sensing underwater, which indicates rich flow information encoded in the whisker array signals. In this study, we aim to investigate how seal whiskers encode information of vortices through flow-whisker interaction (FWI) and how the information obtained by whisker arrays can be used to infer the source of the vortices. We use a one-way FWI simulation to generate large-scale signal data of whisker arrays subjected to vortical wakes of various combinations of cylindrical objects. A deep learning (DL) model is trained on the simulated signals to infer the shape and location of upstream objects. The model is then tested on a smaller-scale experimental dataset. The DL model shows high accuracy for both simulated and experimental data, demonstrating the differentiability of the source using the whisker array signals. In addition to basic information such as frequency and amplitude, the whisker encodes the timing of vortex passing by through the stick-slip mechanism. It is hypothesized that the spatial distribution of these pieces of information contributes to a unique perception of the upstream vortex generator.