Adaptive aggregation of modular fuzzy control

In this paper, a modular intelligent systems design is proposed for fuzzy logic control (FLC) of multi-input and multi-objective systems. Often it's difficult to design fuzzy IF-THEN rules using conventional inference methods when the dimension of the system is large. Furthermore, single input FLC modules are derived for each antecedent in the conventional fuzzy IF-THEN rule. It has shown that the decomposed modular controllers are comparable to nonmodular controllers with the choice of proper weighting coefficients in the final aggregated control action. Moreover, an adaptive aggregation scheme is introduced to tune the weighting coefficients such that the overall performance of the system is satisfied. The advantage of this approach requires less fuzzy linguistic parameters to be adjusted than the traditional nonmodular fuzzy approaches. The proposed design scheme is applied to the cart-pole balancing problem, a multi-input and multi-objective system. The simulation results show the performance of the modular design to be effective for this system.