Title :
A new RBF neural network control strategy based on new object function
Author :
Wan, Ya-Min ; Wang, Sun-an ; Du, Hai-feng
Author_Institution :
Dept. of Mechatronics engineering, Xi´´an Jiaotong Univ., China
Abstract :
The general object function of a neural network (NN)´s learning algorithm is a function of error. We know that the phase space can show the performance of the control system. When the area surrounded by the phase track in the phase space is smaller, the performance of the system is better. So the integrated object function based on the phase space is proposed in the paper. The object function considers synthetically error and its differential coefficient. The new control strategy of a radial basis function (RBF) NN based on this object function is presented, and a new learning algorithm is derived. Experiment results show that the new control strategy can follow the desired output well and converge quickly. It is practical and effective for different complex systems.
Keywords :
learning (artificial intelligence); neurocontrollers; radial basis function networks; RBF neural network control strategy; control structure; learning algorithm; object function; phase space; phase track; radial basis function network; Artificial intelligence; Control systems; Electronic mail; Learning; Mathematical model; Mechatronics; Neural networks; Radial basis function networks; Uncertainty; Wide area networks;
Conference_Titel :
Machine Learning and Cybernetics, 2002. Proceedings. 2002 International Conference on
Print_ISBN :
0-7803-7508-4
DOI :
10.1109/ICMLC.2002.1174495