Adaptive Controller Based on Experimental Data‐Consistent With Strict Convergence Learning Law for Fuzzy‐Rules Network

ABSTRACT

This article introduces a strictly convergent—in the sense of Semi‐Global Uniform Ultimate Boundedness (SGUUB)—model‐free adaptive control (MFAC) strategy for unidentified discrete‐time nonlinear systems. The central novelty lies in a data‐driven synthesis, where the control structure is formulated entirely from experimental input‐output profiles, thereby removing any reliance on analytical system models. Utilizing a Multi‐Input Fuzzy Rules Emulated Network (MiFREN) with a condensed parameter footprint, this method effortlessly handles non‐smooth nonlinear behaviors like severe friction and dead‐zones, bypassing the need for state observers or extra compensatory terms. A robust learning law, designed for strict convergence to a small bounded residual set, delivers formal SGUUB stability guarantees while simultaneously mitigating the harmful high‐frequency noise often seen in traditional adaptive signals. Extensive hardware‐in‐the‐loop testing on a DC motor apparatus highlights a dual advantage: the proposed controller secures exceptional tracking accuracy and refines the frequency spectrum of the applied control effort. By eradicating damaging mechanical chattering, the approach lessens hardware fatigue and noticeably prolongs the lifespan of the physical components.