Abstract: A systematic study of one-, two- and three-input fuzzy inferences has been made for fuzzy PID qontrollers. We propose four criteria to evaluate the functional behaviors of the controllers, namely, control-action composition, coupling influence, gain dependency and rule growth. Based on a simple analysis, we conclude that the Mamdani's two-input fuzzy controller suffers from the difficulties according to the criteria. A novel one-input fuzzy controller is proposed which consists of a "ID-3D" fuzzy inference to generate three independent (proportional, integral and derivative)control actions. This system has shown the best inference structure with respect to the four functional criteria in comparison with two- and three-input fuzzy controllers.