High-speed Integer Operations in the Fuzzy Consequent Part and the Defuzzification Stage for Intelligent Systems

摘要

In high-speed fuzzy control systems applied to intelligent systems such as robot control, one o f the most important problems is the improvement of the execution speed of the fuzzy inference. In particular, it is more important to have high -speed operations in the consequent part and the defuzzification stage. To improve the speedup of fuzzy controllers for intelligent systems, this paper presents an integer line mapping algorithm to convert (0, 1) real values of the fuzzy membership functions in the consequent part to a 400 × 30 grid of integer values. In addition, this paper presents a method of eliminating the unnecessary operations of the zero items in the defuzzification stage. With this representation, a center of gravity method can be implemented with only integer additions and one integer division. The proposed system is analyzed in the air conditioner control system for execution speed and COG, and applied to the truck backer-upper control system. The proposed system shows a significant increase in speed as compared with conventional methods with minimal error; simulations indicate a speedup of an order of magnitude. This system can be applied to real-time high- speed intelligent systems such as robot arm control.