A modified defuzzifier for control of the inverted pendulum using learning

The paper introduces a novel defuzzification procedure to segment the input domains and assign selectable gain factors. With proper factor values in the defuzzifier, a fuzzy logic controller can speed-up the transient response without affecting stability. In addition, an instability due to inappropriate control rules can be accommodated. Also, the proposed defuzzifier can improve the system performance by using a genetic algorithm to learn the gain factors.