Modeling of Pam-based Position Servosystem Static Hysteresis Using Anfis

This paper deals with static hysteresis modeling for the PAM-based (pneumatic artificial muscles) position servosystem by means of an adaptive neuro-fuzzy inference system. The hysteretic behavior is expressed in arm position-pressure difference dependence where the output variable (arm position) depends on the history of input variable (pressure difference). In this way, a first order (until input first reversal) curve and major loop of hysteresis are modelled, with separation technique being chosen for modeling. The function itself is approximated using adaptive neuro-fuzzy inference system which proved very effective in terms of approximation error even in case of sparse training sets. All the measurements and modelling processes are carried out in Matlab/Simulink with Fuzzy Logic Toolbox.