Generación y optimización de controladores difusos utilizando el modelo NEFCON
Generation and Optimization of Fuzzy Controllers Using the NEFCON Model
Erik V. Cuevas Jiménez1,2, Daniel Zaldívar Navarro1,2, Marco Pérez Cisneros1 y Ernesto Tapia Rodríguez2
1 Departamento de Ciencias computacionales, Universidad de Guadalajara, CUCEI Av. Revolución 1500, Guadalajara, Jal, México. E–mail: erik.cuevas@cucei.udg.mx, daniel.zaldivar@cucei.udg.mx, marco.perez@cucei.udg.mx
2 Institut fur Informatik, Freie Universität Berlin Takustr. 9, Berlin, Alemania tapia@inf.fu–berlin.de
]]>Artículo recibido en Enero 07, 2008.
Aceptado en Marzo 26, 2009.
Resumen
El diseño de algoritmos que operen sobre plantas con dinámicas no modeladas aún representa un reto en el área de control automático. Una solución podría ser el uso de algoritmos capaces de aprender en tiempo real mediante la interacción directa con la planta. El modelo NEFCON, permite construir la estructura de un controlador difuso del tipo Mamdani capaz de aprender las reglas y adaptar los conjuntos difusos. La principal ventaja del modelo NEFCON respecto a otros enfoques de aprendizaje, es que su diseño se reduce a expresar la calidad del error actual de la planta a controlar. Sin embargo, una desventaja del modelo NEFCON es la pobre exploración de los estados de la planta durante el aprendizaje, lo cual hace imposible su aplicación para sistemas dinámicos no lineales. En este trabajo se propone la adición de ruido Gaussiano a las variables de estado de la planta, con el objetivo de asegurar una exploración amplia de los estados, facilitando la convergencia del algoritmo de aprendizaje, cuando se aplica a sistemas no lineales. En particular, se muestra la efectividad de la propuesta en el control del sistema dinámico de la "pelota y el balancín" (Ball and Beam)
Palabras clave: Sistemas de control adaptativos, sistemas de control por aprendizaje, control inteligente, control no lineal.
Abstract
The design of algorithms that operate on un–modeled dynamics plants still represents a challenge in automatic control area. A solution could be the use of algorithms able to learn in real time by direct interaction with the plant. NEFCON, allows to build a Mamdani fuzzy controller able to learn rules and adapt the fuzzy sets. The main advantage of NEFCON compared with other learning approaches, is that its design express the current error state of the plant to be controlled. However, a disadvantage of NEFCON is its poor exploration of the states of the plant during the learning; disable its application on nonlinear dynamic systems. In this work the addition of Gaussian noise to the states of the plant is proposed with the objective to assure a wide exploration of the states, simplifying the convergence, when it is applied to nonlinear systems. In particular, the effectiveness of our proposal is shown in the control of the "ball and beam" dynamic system.
]]> Keywords: Adaptive control systems, learning control systems, intelligent control, nonlinear control.
DESCARGAR ARTÍCULO EN FORMATO PDF
Referencias
1. Brown, M. & Harris, C.J. (1994). Neurofuzzy Adaptive Modelling and Control. Hemel Hempstead: Prentice Hall. [ Links ]
2. Cuevas, E. & Zaldívar, D. (2006). Sistemas de Control Neurodifuso. Göttingen: Cuvillier Verlag. [ Links ]
3. Dominguez, J. A., Damper, R. I. & Harris, C. J. (2004). Adaptive neurofuzzy control of a robotic gripper with online machine learning. Robotics and Autonomous Systems, 48 (2), 93–110. [ Links ]
4. Gonzalez, M. A. & Tang, Y. (2007). A new recurrent neurofuzzy network for identification of dynamic systems. Fuzzy Sets and Systems, 158 (10), 1023–1035. [ Links ]
5. Gupta, M.M. & Sinha, N.K. (1999). Soft Computing and Intelligent Systems: Theory and Applications. Amsterdam: Elsevier. [ Links ]
6. Hangos, K.M., Lakner, R. & Gerzson, M. (2001). Intelligent Control Systems: An Introduction with Examples. Netherlands: Kluwer Academic Publishers. [ Links ]
7. Harris, C., Hong, X. & Gan, Q. (2002). Adaptive Modelling, Estimation and Fusion from Data: A Neurofuzzy Approach. Berlin: Springer–Verlag. [ Links ]
8. Haykin, S. (1999). Neural Networks: A Comprehensive Foundation. Upper Saddle River, NJ: Prentice Hall. [ Links ]
9. Jang, J.–S.R. (1993). ANFIS: Adaptive–Network–based Fuzzy Inference System, IEEE Transactions on Systems, Man, and Cybernetics 23 (3), 665–685 [ Links ]
10. Jang, J–S.R., Sun, C–T. & Mizutani, E. (1997). Neuro–Fuzzy Modeling and Soft Computing: A Computational Approach to Learning and Machine Intelligence, Englewood Cliffs. NJ: Prentice–Hall. [ Links ]
11. Lon–Chen, H. & Hung–Yuan, C. (2007). Decoupled control using neural network–based sliding–mode controller for nolinear systems, Expert Systems with Application, 32 (4), 1168–1182. [ Links ]
12. Moon, G. & Jin, S. (2002). Universal approximation by hierarchical fuzzy system with constraints on the fuzzy rule, Fuzzy Sets and Systems, 130 (2), 175–188. [ Links ]
13. Nauck, D. & Kruse, R. (1992). A Neural–Fuzzy Controller Learning by Fuzzy Error Propagation, North American Fuzzy Logic Processing Society, Puerto Vallarta, Mexico, 388–397. [ Links ]
]]>14. Nauck, D. (1994). Fuzzy Perceptron as a Generic Model for Neuro–Fuzzy Approaches, Fuzzy Systeme'94, 2nd Generación y Optimación de Controladores Difusos Utilizando el Modelo NEFCON 131 GI–Workshop Siemens Corporation, Munich, Germany. 32–37. [ Links ]
15. Nauck, D., Kruse, R. & Klawonn, F. (1997). Foundations on Neuro–Fuzzy Systems, Chichester: Wiley. [ Links ]
16. Nauck, D., Klawonn, F. & Kruse, R. (1996). Neuronale Netze und Fuzzy–Systeme (2nd extented ed.), Wiesbaden: Vieweg Verlag. [ Links ]
17. Nürnberger, A., Nauck, D. & Kruse, R. (1999). Neuro–Fuzzy Control Based on the NEFCON–Model (2007), retrieved from http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.28.6827 [ Links ]
18. Perez–Cisneros, M.A. & Wellstead, P. (2004). Sistema de Balancín y Pelota: Principios Básicos, (notas de aplicación). México: División de electrónica y computación, Universidad de Guadalajara. [ Links ]
19. Rousseau, D. (2005). Constructive action of additive noise in optimal detection, International Journal Bifurcation Chaos,15 (9), 2985–2994. [ Links ]
20. Shi–Yuan, C., Fang–Ming, Y. & Huang–Yuan, C. (2002). Decoupled fuzzy controller design with single–input fuzzy logic. Fuzzy Sets and Systems, 129 (3), 335–342. [ Links ]
21. Sung–Kwun, O., Seok–Beom, R. & Pedrycz, W. (2007). IG–based genetically optimized fuzzy polynomial neural networks with fuzzy set–based polynomial neurons, Neurocomputing, 70 (16–18), 2783–2798. [ Links ]
22. Sutton, R., & Barto, A. G.(1998). Reinforcement Learning: An Introduction. Cambridge: MIT Press. [ Links ]
23. Yen, J., & Langari, R. (1999). Fuzzy Logic, Intelligence, Control, and information, Upper Saddle River: Prentice–Hall. [ Links ]
]]>