Optimization of a Fuzzy PI Controller using Reinforcement Learning
Hamid Boubertakh, Pierre Yves Glorennec · 2006
This paper proposes a methodology for fine tuning of the conclusion part of fuzzy proportional-integral controllers (FPIC), using both a reinforcement learning method and all the available knowledge on the process under control. Membership functions on the error domain and rule conclusions are easily defined. Therefore only the conclusion part have to be tuned