Weight Design Policy for Two-Step Gain Optimization Method of Model Free Water-Hydraulic Muscles Controller Based on Ultra-Local Model by VRFT

Abstract

A McKibben-type artificial muscle has several advantages for practical applications, however, it is challenging to control it with high-precision due to its strong hysteresis characteristic caused by friction. For this reason, in this paper, a VRFT-MFC that combines Model-Free Control (MFC) and Virtual Reference Feedback Tuning (VRFT) was proposed, which is based on MFC structure, and VRFT optimizes the parameters of MFC with a pair of input-output data to satisfy a stability condition for the entire system. However, this method includes weights design that strongly affects control performance, and their tunings are time-consuming routines. To solve these problems, we propose a design policy for the weights. It is confirmed that the effectiveness and optimality of the proposed policy results for water-hydraulic McKibben-type artificial muscles in experiments.