Adaptive Fuzzy Rules Emulated Networks Controller for 7-DOF Robotic Arm with Time Varying Learning Rate - Developing Concepts in Applied Intelligence - page 106

Adaptive Fuzzy Rules Emulated Networks
Controller for 7-DOF Robotic Arm with Time
Varying Learning Rate

C. Treesatayapun

Department of Robotic and Manufacturing, CINVESTAV-Saltillo,
Ramos Arizpe, 25900, Mexico

Abstract. This article presents an adaptive controller based on fuzzy
rules emulated network (FREN) with the proposed on-line learning algo-
rithm. The human knowledge about the unknown plant, 7- DOF robotic
arm in this case, is transferred to be if-then rules for setting the network
structure. All adjustable parameters are tuned by the on-line learning
mechanism with time varying step size or learning rate. The main theo-
rem is introduced to improve the system performance and stabilization
through the variation of learning rate. Experimental system based on
Mitsubishi PA-10 is demonstrated the control algorithm validation.

Keywords: Discrete-time; adaptive control; neuro-fuzzy; 7-DOF robotic
arm.

1 Introduction

To design the controller for robotic manipulators, dynamic and kinematic models
usually are the normal requirements even those controllers has been constructed
by artificial intelligence tools such as neural networks (NNs) and fuzzy logic sys-
tems [1,2]. The system performance is commonly related on the model accuracy
or approximated model in the case of using NNs to estimate robotic models. In
this work, on the other hand, a direct adaptive controller based on fuzzy rules em-
ulated networks [3] (FRENs) is introduced with out any requirement of robotic
system modeling. According to an on-line learning, the close-loop performance
can be guaranteed by using time varying learning rate.
Recently, to ensure the system performance small learning rates or step sizes are
often used in gradient search methods because of the system stability but the small
step size can slow down the reaching solution [5]. In [6], the suitable learning rate
which can guarantee the system stability has been widely discussed but the using
limit or operation range has been considered because of fixed variable learning rate.

2 Robotic System with Non-ane Discrete-Time Systems

According to design the controller with out any requirement of robotic system
modeling, this section, we consider these robotic systems as a class of non-ane
formulation as follows: