Title :
Reinforcement-driven adaptation of control relations
Author :
Jacobsen, Hans Arno ; Weisbrod, Joachim
Author_Institution :
Berkeley Initiative in Soft Comput., California Univ., Berkeley, CA, USA
Abstract :
The conceptual framework of a hybrid control system architecture is briefly discussed. It employs neural and fuzzy techniques side-by-side using each one for the task to which it is best suited. Our main interest is with the adaptation of the fuzzy control knowledge. The adaptation algorithm is based on reinforcement signals and directly optimizes the global fuzzy relation representing the complete knowledge base. The new approach is experimentally evaluated
Keywords :
fuzzy control; fuzzy neural nets; intelligent control; neurocontrollers; unsupervised learning; complete knowledge base; control relations; fuzzy control knowledge; fuzzy techniques; global fuzzy relation; hybrid control system architecture; neural techniques; reinforcement signals; reinforcement-driven adaptation; Availability; Control systems; Fuzzy control; Fuzzy sets; Fuzzy systems; Humans; Jacobian matrices; Knowledge based systems; Neural networks; Supervised learning;
Conference_Titel :
Fuzzy Information Processing Society, 1996. NAFIPS., 1996 Biennial Conference of the North American
Conference_Location :
Berkeley, CA
Print_ISBN :
0-7803-3225-3
DOI :
10.1109/NAFIPS.1996.534782
