Soft Proprioception in the Frequency Domain via Tube Resonance

Soft robots require proprioceptive sensors capable of measuring their complex body deformations. We introduce a novel approach to soft proprioception inspired by the deterministic acoustic responses of wind instruments, where standing-wave resonance generates distinctive acoustic signals sensitive to the size and location of tone holes. This proprioceptive sensor can produce distinct frequency-domain signatures measurable via a microphone, which correlate with the magnitude and location of the tube's longitudinal curvatures. We demonstrate that the acoustic signals can be interpreted through the sensor's acoustic boundary conditions. In addition, we show a prototype sensor that achieved reliable real-time classification of 9 different tube bending poses, with a weighted recall of 0.984 using a nonlinear random forest model, while a linear logistic regressor achieved a weighted recall of 0.889. This musical-instrument-inspired design offers a low-cost and computationally efficient approach to proprioception that can be potentially adapted to a wide range of forms and applications.