Linear vs. Nonlinear Modeling of Continuum Robotic Arms Using Data-Driven Method

Abstract

Dynamics modeling of continuum robotic arms is of great importance due to the highly nonlinear characteristics, uncertain and complex structure, and the inherent underactuation. This affects further usage in various aspects, including inverse kinematics, trajectory generation, control and optimization. In this paper, a modelling approach is proposed through the use of data-driven identification by linear and nonlinear models known as ARX (autoregressive with exogenous terms) and NARX (nonlinear autoregressive with exogenous terms) models. The unknown parameters in the ARX model are the system parameters; while the structure is known. However, for NARX model, the whole structure is considered to be unknown. These two structures are used to model a single-section continuum robotic arm, and the results are compared. Finally, the advantages and disadvantages of them are discussed.