Fuzzy controller synthesis for an inverted pendulum system

Abstract A frequently discussed issue in the use of fuzzy systems for control design is related to the ad hoc nature by which controller synthesis is performed, where incorporation of the designer's knowledge into the synthesis procedure is often not straightforward. This paper describes a controller synthesis procedure based on the idea of expanding the usable region of a linear control design technique. Comparative analyses are carried out via implementation of swing-up and balancing control for a rotational inverted pendulum system. For the complete control problem, the paper focuses on three aspects of fuzzy control in the overall control design: direct, supervisory , and auto-tuning fuzzy control. For swing-up control, an energy pumping strategy , enhanced by introducing a fuzzy supervisory mechanism, is utilized. For balancing control, an LQR-based linear control strategy, valid in the “linear region” of operation of the pendulum, is generalized to a nonlinear direct fuzzy control design. For improved performance when modeling uncertainties affect controller performance, a new scheme is developed, again based on the LQR design, which is a simple yet effective auto-tuned fuzzy controller, having the capability to vary its resolution on-line. All controllers are tested on an experimental apparatus, and performance comparisons are drawn.