Compensating dynamics of impedance haptic devices using Neural Networks

This paper presents a Neural Network approach to compensate dynamic terms, friction force in particular, of a four degree of freedom haptic device manipulator similar to commercial one's that are on the market, which is controlled in impedance. The friction force model is analyzed using a general compensation method after which a trained Multi-Layer Neural Network is introduced in order to obtain a more accurate friction approximation for cancelling out this term from dynamics so that the movement of the device feels free and unconstraint.