Fuzzy and Neural Controllers for a Pneumatic Actuator

Abstract

There is a great diversity of ways to use fuzzy inference in robot control systems, either in the place where it is applied in the control scheme or in the form or type of inference algorithms used. On the other hand, artificial neural networks ability to simulate nonlinear systems is used in different researches in order to develop automated control systems of industrial processes. In these applications of neural networks, there are two important steps: system identification (development of neural process model) and development of control (definition of neural control structure). In this paper we present some modelling applications, which uses fuzzy and neural controllers, developed on a pneumatic actuator containing a force and a position sensor, which can be used for robotic grinding operations. Following the simulation one of the algorithms was tested on an experimental setup. The paper also presents the development of a NARMA-L2 neural controller for a pneumatic actuator using position feedback. The structure had been trained and validated, obtaining good results.