A new approach for modelling and identification of the pneumatic artificial muscle manipulator based on recurrent neural networks

Abstract

Pneumatic artificial muscles (PAMs) are widely used in the various fields of medical robots and other industrial applications. As a powerful tool of intelligent control, neural networks nowadays are applied effectively to model, identify, and control highly non-linear systems, including the PAM manipulator. In the current paper a prototype PAM manipulator is modelled through recurrent neural networks (NN) modelling and identification based on experimental input-output training data. The proposed incremental back-propagation (INCBP) algorithm, which yields faster convergence than a conventional back-propagation (BP) algorithm, is applied to train the neural networks. The realization of the INCBP algorithm is given. An evaluation is carried out for different non-linear NN auto-regressive with exogenous input (NNARX) models of a PAM manipulator, using recurrent NN with various input nodes as well as various hidden layer nodes. For the first time, the non-linear NNARX model scheme of the prototype PAM manipulator has been investigated. The results show that the non-linear NNARX model trained by INCBP yields better performance and higher accuracy than the traditional linear ARX model. These results can be applied in modelling, identifying, and controlling not only the PAM manipulator, but also other highly non-linear systems.