Sensorless Proprioception in Multi-DoF Dielectric Elastomer Soft Robots via System-Level Self-Sensing

Abstract

Proprioception in soft robots remains an ongoing challenge, owing to practical issues such as tight sensor integration while maintaining mechanical compliance. Soft robots based on dielectric elastomer (DE) technology could provide a compelling answer to these challenges, due to the inherent flexibility and compliance of such transducers as well as their ability to simultaneously work as actuators and sensors, i.e., self-sensing. In this work, we propose a novel real-time self-sensing scheme for DE soft robotic systems. By combining an actuator-level recursive least squares identification with an extended Kalman filter, our architecture provides an estimation of the mechanical state of the structure without requiring additional electromechanical sensors, solely relying on electrical measurements performed on the DEs during high voltage actuation. Experimental validation, conducted on a DE soft robot prototype, reveals that the proposed solution reconstructs the system state during actuation in a robust and accurate way, and under various external loading conditions.