Comparison of Least Square Identification Schemes for a Model-Free Fuzzy Adaptive Controller

Fuzzy adaptive controllers can guarantee tight control performance when the plant under control is uncertain and non-linear. Adaptive algorithm(s) which is an integral component of the control strategy plays a significant role in determining the control performance. In this work, the performance of a model-free fuzzy adaptive controller (MFFAC) is compared for two different variations of the adaptive algorithm. The adaptive fuzzy controller is modeled as a Takagi–Sugeno model. The MFFAC develops an online inverse model of the plant. Adaptive algorithms based on least mean square (LMS) are extensively used for online learning. In this paper two variations of the LMS algorithm are compared. The two algorithms are normalized least mean square and fuzzy least mean square. Both the adaptive schemes were tested for different parameters of the adaptive fuzzy controller. The aim was to investigate the robustness of the controller and the rate of convergence of the controller parameters when the plant is subjected to noise and disturbances.