Support vector machine networks for friction modeling

This paper presents a novel model-free approach for modeling friction for servo-motion systems. The proposed approach uses the support vector machine networks to parameterize the static friction mapping. The procedure of constructing such networks from a finite amount of training (sampling) data is developed based on support vector machine regression (SVMR). The validity of the proposed approach has been experimentally verified.