Elastic Adaptive Fuzzy Logic Controller
Ognjen Kuljača, Frank L. Lewis, Jyotirmay Gadewadikar, Krunoslav Horvat · Journal of Communication and Computer · 2010
Abstract: The authors present an adaptive fuzzy logic controller that includes on-line tuning of the membership function centroids and spreads, as well as an additional parameter, the “elasticity”. The notion of elasticity, taken from economics, allows weighting of the relative importance of each individual factor in a rule. A novel membership function is introduced here that satisfies certain important properties, including an independent reaction when the elasticity is tuned. We provide architecture for an elastic fuzzy logic adaptive controller, using certain on-line tuning laws that are an enhanced form of backpropagation tuning. It is shown by simulation that the fuzzy logic controller that includes the new membership function and tuning of elasticity outperforms the standard adaptive fuz zy logic controller based on Gaussian membership functions. Key words: Adaptive, elastic, fuzzy logic, control. 1. Introduction the literature are Guassian membership functions. Adaptive fuzzy logic (AFL) systems are becoming more and more popular in control systems due to the ability to select initial membership functions (MFs) based on experience and intuition, and the ability to tune the MFs to learn about the unknown dynamics of the system. Due to their approximation property, fuzzy logic (FL) systems can be tuned to estimate the unknown functions in dynamical systems and to reject disturbances. By now, proofs of the stability and performance analysis of FL systems have been provided by a variety of researchers (Refs. [1-10] and the others). The most common adaptive fuzzy logic controllers (FLCs) membership functions described in