Neural Network based Pulley Friction Compensation for Tension Control of a Cable-Driven Parallel Robot

Cable driven parallel robots (CDPRs) have several advantages compared to conventional parallel robots such as large workspace, high dynamics, and high payload by virtue of elastic lightweight cable links instead of existing rigid links. However, the CDPRs have a force and position control problem caused by nonlinearity and uncertainty in pulleys and elastic cable. Notably, the friction between pulley and cable brought significant force differences between winch-motor and end- effector. In this paper, we present the pulley friction compensation method to obtain direct tension at the end-effector without utilizing additional force sensor by using an Artificial Neural Network (ANN). The proposed control algorithm which is composed of a PID-controller to track the desired tension and an ANN to estimate unmodeled force differences between winch-motor tension and the end-effector tension. The ANN could estimate the friction in the pulley used for the CDPRs based on the measured datasets of the elastic cable for the given experimental condition. The experimental results showed that pulley friction could be remarkably compensated during single cable tension control.