Optoelectronic analog antecedent and consequent fuzzy inference units using LED arrays and position sensing devices

A novel implementation of analog fuzzy inference architecture with optoelectronic antecedent and consequent operation units is realized. The inference process uses Gaussian-like membership functions and a product-sum gravity method. Antecedent membership functions and the grade evaluations are realized by combining LED arrays for sequential light emission and position-sensing devices for center-of-gravity detection. Consequent and defuzzification operations are utilized by an LED array and a position-sensing device. High-speed fuzzy inference of more than several tens of MFLIPS is possible using those units. Usefulness of the system is demonstrated by a realtime standing control of an inverted pendulum.