Pneumatic reservoir computing for sensing soft body: Computational ability of air in tube and its application to posture estimation of soft exoskeleton

Abstract

Sensing and computing the body states of soft-bodied robots require new methods. Recent studies on soft robotics have shown their feasibility to be applied for these purposes; however, they only addressed solid parts and not the behavior of inner fluids in soft robotics. In this study, we investigated the possibility of a framework that can be used to estimate the body state by exploiting air dynamics in tubes connecting chambers. The framework was designed on the basis of the concept of physical reservoir computing. We focused on a case with a single tube connection. A benchmark task emulated a nonlinear system and was evaluated by simulation. The results showed that the computational ability depended on the inner diameter and length of the tube and can be increased by selecting a suitable diameter and length. We physically implemented the framework for the posture estimation of a soft exoskeleton using pneumatic artificial rubber muscles (PARMs) as the connected chambers and evaluated the accuracy of estimation of a thigh angle. The estimation accuracy showed a similar trend as a function of the tube properties as that observed in the simulation. The framework can exploit the dynamics of air in a tube and may be useful for the state estimation of soft-bodied robots.