Designing a Muscle like System Based on PID Controller and Tuned by Neural Network

Author

Serhan, Hayssam J. ; Nasr, Chaïban G. ; Henaff, Patrick

Author_Institution

Lebanese Univ., Beirut

fYear

0

fDate

0-0 0

Firstpage

4991

Lastpage

4998

Abstract

This paper presents a study on a muscle like system based on a PID controller tuned by a neural network. The approach is based on a non linear muscle model using system identification based on a NNARX (neural network autoregressive exogenous) structure. This model is used in a special configuration of an MLP in order to let the output of the closed loop formed by the motor and controller to follow that of this non linear muscle model. Two structures are compared and the robustness of the approach is tested with different models of DC motors.

Keywords

autoregressive processes; closed loop systems; control system synthesis; identification; learning systems; legged locomotion; motion control; neurocontrollers; nonlinear control systems; robust control; three-term control; PID controller; closed loop system; muscle like system design; neural network autoregressive exogenous structure; nonlinear muscle model; robustness; system identification; walking robot; Control systems; DC motors; Humans; Leg; Legged locomotion; Linear systems; Muscles; Neural networks; System identification; Three-term control;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks, 2006. IJCNN '06. International Joint Conference on

Conference_Location

Vancouver, BC

Print_ISBN

0-7803-9490-9

Type

conf

DOI

10.1109/IJCNN.2006.247203

Filename

1716794