離散時間反覆學習控制器之設計

Abstract

摘要 近來，反覆式學習控制器於學習能力相關研究領域上，已展現良好的應用能力；但在反覆式學習控制器設計方法的討論中，對於初始狀態誤差的研究，仍需更多努力的投入。 本論文提出新的反覆式學習控制器設計之觀點來設計控制器，以增進初始狀態誤差的收斂性。而所設計的控制器效能是以電腦模擬來進行驗證，模擬結果與理論推導的分析一致，證明新設計的反覆式學習控制器皆能達成前述的設計要求；此外，論文中亦呈現一順向模糊類神經網路利用該反覆式學習控制器的運算結果，來訓練網路並完成其網路連結的調整，模擬結果也發現可以獲得良好的學習效果。 總而言之，設計反覆式學習控制器的觀點與模糊類神經網路的應用，可以視為此二領域整合研究的另一起點。

ABSTRACT ILC ( Iterative Learning Controllers ) are recently showing good promise for applications in learning related research areas. However, the initial state error problem is still an important research field that needs more attention in ILC design methodology. This thesis proposes new ILC design concepts to improve convergence for the initial state error problem. The effectiveness of the proposed controllers is demonstrated by some simulation results. The theoretical analysis as well as the simulation results to verify the effectiveness in improving error convergence of the new designed ILC controllers are described in this thesis. Besides, an FNN ( Feedforward Fuzzy Neural Network ) receives the iterative information from the ILC to train. And the improved competence in learning is achieved also. In conclusion, the proposed ILC controllers and FNN can be viewed as an another starting point in the research of the integration of both.