Static Elasticity Compensation via Recursive Artificial Neural Network for Long-Reach Cable-Driven Hyper-Redundant Manipulators

Abstract

In recent years, long-reach cable-driven hyper-redundant manipulators have been spreading into the field of robotics inspection, thanks to their ability to move through narrow spaces and harsh environments. One of the key challenges in achieving precise motions using these robots is compensation for static deformation of the cables. In this context, a recursive algorithm for online training of neural networks that can compensate for such phenomena has been implemented. A Model-Based compensation method and compensation performed via an Artificial Neural Network have been used as baseline comparisons to highlight the improvements provided by Recursive Artificial Neural Network developed in this paper. A long-reach cable-driven hyper-redundant manipulator has been used as the test bench to evaluate the performances of the recursive algorithm. The experimental results have shown a notable improvement in terms of reduction of the tracking error.