SELF-TRAINING TYPE NEURO-FUZZY NETWORK ARCHITECTURE AND ELECTRONIC CONTROL METHOD BY SELF-TRAINING FOR ELECTRONIC EQUIPMENT

摘要

PROBLEM TO BE SOLVED: To obtain a new neuro-fuzzy integration architecture capable of realizing on-line self-training. SOLUTION: This architecture is provided with a fuzzy type microcontroller 11 exclusive for fuzzy rule calculation and integrated on a semiconductor together with a non-volatile memory 9 so as to be monolithic, and the same IC is provided with a microprocessor 5, a volatile memory unit 2, and an arbiter block 3 connected to a bus 4 connecting the fuzzy microcontroller 11 and the microprocessor 5 and the volatile memory unit 2. The arbiter block 3 controls access to the volatile memory unit 2 by the microcontroller 5 or the fuzzy microcontroller 11. Then, an additional fuzzy co-processor 6 for fuzzy logical arithmetic processing is connected between the fuzzy microcontroller 11 and the microprocessor 5.