Data-Driven Modeling of Soft Robots via Koopman Operator Theory

Abstract

To address the problem of accurate modeling of soft robots, a data-driven modeling method based on the Koopman operator is proposed. The method aims to achieve global linearization of highly nonlinear dynamical systems in an infinite-dimensional lifted state space. By employing the Extended Dynamic Mode Decomposition (EDMD) algorithm, the infinite-dimensional linear Koopman operator is approximated, and both linear and nonlinear models of the soft robot dynamical system are established. The results demonstrate that the two Koopman-based models achieve higher accuracy than conventional state-space models, and the Koopman nonlinear model exhibits the best modeling and prediction performance.