Fuzzy linear regression for contact identification

Abstract

In order to obtain an autonomous and intelligent system dealing with the uncertainties occuring in force controlled tasks, the determination of the parameters pertaining to the contact situation between the robot end-effector and the object is required. The paper describes a methodology based on fuzzy linear regression presenting two kinds of fuzzy linear models. The inputs are the force and velocity measurements. The results are compared with statistical regression. Moreover, the influence of initial data modelling in terms of extent as well the influence of the optimization constraint in the regression model are investigated.