Reduced-Order Modeling of a Soft Anthropomorphic Finger for Piano Keystrokes

The soft-bodied robotic hand provides a new paradigm for developing intelligent robotic systems with omnidirectional flexibility. This paper presents a novel approach to modeling the behavior of a soft anthropomorphic finger used in piano keystrokes. The proposed reduced-order model aims to capture the essential dynamics and characteristics of the finger's motion during piano playing. By leveraging the particle jamming technique, we design and fabricate a variable stiffness finger to achieve soft-bodied touch on a piano. The interaction between the soft finger and the piano keys is approximated using a high-dimensional linear time-invariant system, with state-space model parameters identified. To facilitate rapid computation and efficient keystroke control, we apply frequency-limited balanced truncation for model order reduction in the context of single-note piano playing. Our experimental results demonstrate the effectiveness of frequency-limited balanced truncation in compressing the high-dimensional state-space model while accurately representing the finger's kinematics during piano keystrokes under various stiffness conditions.