Iterative learning fuzzy gain scheduler

Abstract

Humans learn from experience and from repeating a task again and again. This paper uses this human capability to iteratively adjust fuzzy membership functions. These learnt membership functions are used to schedule the gain of a conventional proportional controller. The adjustment of the membership functions help achieve the design requirements of steady state error and percentage overshoot while the scheduling of the proportional gain gives us an adaptive controller. Simulation and experimental results demonstrate that this simple yet robust approach can be applied as an alternative to proportional-integral- derivative (PID) controllers, which are extensively used for various applications. © 2008 Springer-Verlag.