|
|||
An online Learning
Algorithm for Neurocontroller based on a
Fuzzy Estimate of the Control Error
B.Boutamina, K.Belarbi, S. Filali
Laboratoire d’automatique et de robotique, Faculty of engineering, University of Constantine,
Route de
Ain el Bey, Constantine, 25000, Algeria
Abstract: An online learning strategy for neurocontrollers based on minimisation of the control error is proposed. Since this latter is unknown a fuzzy system infers its estimate from the plant output error and its variation. This estimate is then used for weight updating in an online backpropagation algorithm. This procedure affects only the step size of the updating law if the estimate has the correct sign. The rule basis is derived from heuristic analysis. The algorithm is applied in simulation for the control two nonlinear systems. The results show that learning is fast and depends on the learning rate. Keywords: Neural network control, fuzzy systems, online learning, backpropagation
@ARTICLE{P1110617001,
AUTHOR = {B.Boutamina and
K.Belarbi and S. Filali},
TITLE = {An online Learning Algorithm for Neurocontroller based on a Fuzzy Estimate of the Control Error}, JOURNAL = {ICGST International Journal on Automatic Control and Systems Engineering, ACSE}, YEAR = {2006}, VOLUME = {06}, ISSUE = {II}, PAGES = {27--31} }
|
|||
|
