通常在設計模糊控制器，包含獲得資料、定義控制結構、定義規則庫及其他的控制參數時，都是相當費時的。目前，有一項重要的議題就是如何減少相關模糊規則的數目，以符合計算上的需求，階層式模糊控制系統的想法因此被提出。可是，在階層化的中間階層裡，相關模糊規則可能只有些許的物理意義而導致難以去控制。而且這種現象在越多階層越明顯。 為了解決這中間階層沒有物理意義的輸出變數，本文提出一個新型式的規則庫對映方案，來求得相關規則而不必考慮其物理意義。如此，所有的規則不必再重新設計，一樣可以達到減少規則數目，而且，在多層架構中，這種對映法一樣有效。 利用電腦模擬來證實本文所提出之方法的可行性，以及說明整個設計過程。再則，對實際系統之實驗，例如倒三角體，輔以基因演算法，更加驗證了這個設計方法的有效性。由這些模擬及實驗結果顯示，本文提出的方法確實提供有效之途徑以設計階層化模糊控制系統。 The design of fuzzy controllers is commonly a time-consuming activity involving knowledge acquisition, definition of the controller structure, definition of rules, and other controller parameters. At present, one of the important issues in fuzzy logic systems is how to reduce the number of involved rules and their corresponding computation requirements. The idea of hierarchical fuzzy systems (HFSs) has been reported. But, the involved fuzzy rules in the middle of the hierarchical structure have little physical meaning and consequently are hard to design. This phenomenon becomes prominent as the number of layers grows larger in an HFS. To overcome the problem that intermediate outputs have nothing to do with the physical variables, this thesis propose a new kind of mapping rule base scheme to get the rule base of HFS without the physical meaning. As a result, all of the rule bases of fuzzy logic units (FLUs) don’t design again and we can reduce the number of involved rules. In many layers, the mapping rule is useful, equivalently. The several simulations on computer are given to confirm the correctness and to illustrate design procedures. Moreover, Experiments on a practical system, such as an inverted wedge system, assisted with genetic algorithm, verify the effectiveness of the proposed methods. Judging from simulative and experimental results, the methods described in this provide efficient approaches to design HFSs.
