A neuro-fuzzy-genetic system for automatic setting of control strategies

The article deals with the design of control systems based on hybrid techniques of computational intelligence. Initially, a neuro-fuzzy system is employed in the control of several plants. The neuro-fuzzy system used here is the NEFCON model, which is capable of learning and optimizing online the rulebase of a Mamdani-type fuzzy controller. The algorithm is based on reinforcement learning that uses a fuzzy measure for the error. Its performances in the control of linear plants of diverse complexity and also of a nonlinear one are evaluated. Results are compared to those obtained through conventional techniques. The main focus of the work is on the development of a new neuro-fuzzy-genetic system, which makes use of genetic algorithms for rule base optimization. The satisfactory results obtained with the two more complex plants show the potential of this hybrid model in the design of control systems