Related Experiment Video
Updated: Aug 16, 2026

A Robotic Platform to Study the Foreflipper of the California Sea Lion
Published on: January 10, 2017
Decoding propulsive wakes using biomimetic seal whiskers
Sanjay Giridharan1, Leixin Ma1
1School for Engineering of Matter, Transport and Energy, Arizona State University, Tempe, AZ 85287, United States of America.
Abstract:
Detecting and tracking swimming animals without disturbing them is a central need in fisheries protection and aquatic biological monitoring, yet optical and acoustic methods remain limited underwater. Passive hydrodynamic sensing offers an alternative: harbour seals can follow hydrodynamic wakes over long distances, and biomimetic whiskers respond measurably to the wakes of upstream bodies. Here we quantify how well four independently varied properties of a propulsive source can be discriminated simultaneously from a single force-instrumented whisker, using towing-tank measurements of ascaled biomimetic seal whisker instrumented with a six-axis force/torque transducer. First, we separate the sensor's intrinsic response from flow-induced forcing in quiescent flow over Reynolds numbers-. The quiescent-flow spectra show that the whisker's self-shedding response is weak and separated from the wake-relevant band. We then expose the sensor to wakes from a flapping NACA 0012 airfoil acrossconditions in which the source frequency, pitching amplitude, angle of attack, and downstream distance are prescribed independently at the source. The measured force response locks to the source frequency, scales with pitching amplitude, varies systematically in the drag-to-lift fluctuation ratio with angle of attack, and decays approximately exponentially with distance, giving four distinct force signatures of the prescribed source state. A one-dimensional convolutional neural network trained on the raw six-channel time series recovers all four attributes simultaneously withaverage test accuracy (mean over five random-seed runs; per-task-) and millisecond-scale CPU inference latency. Frequency and source angle of attack are recovered essentially perfectly, whereas the remaining errors are concentrated in adjacent pitching-amplitude and downstream-distance classes, indicating that the model's failures reflect physically weak separability rather than missing spectral cues. Within the experimentally investigated parameter space, these results establish the force signature of a biomimetic whisker as a compact basis for reading the state of a propulsive wake source.

