小腦模型控制器設計及其於電動運動平台控制之應用

Abstract

本研究主要進行電動運動平台的設計與開發工作，並設計一適應性小腦模型控制系統，用來解決電動運動平台之高複雜度及非線性等控制問題。相較於本實驗室現有的油壓運動平台，本研究所開發的電動運動平台具有低噪音與低污染(無油污漏油的問題)等優點，甚至只需家用110V電源即可驅動馬達的轉動即可使電動平台運作。基於以上這些優點，目前國外製作運動平台的公司，都已經陸續投入心力於開發電動平台以取代現有的油壓平台，而國內外之相關的研究單位對於運動平台的需求，也都將注意力轉移至電動平台開發上。針對此點，本研究分成兩大主題，一為機構硬體設備，另一為控制系統設計。對於機構硬體設備部分，本研究已成功完成電動運動平台的機構設計、加工及驅動系統架設工作;對於控制系統設計部分，由於電動運動平台之動態模型十分複雜，依舊使用傳統的控制理論不易達成良好的控制目的，為解決此問題，我們提出一架構簡單、具快速學習且不需要受控系統動態模型的適應性小腦模型控制器來解決此控制問題。最後經由平台架設與實驗結果發現本研究所提出之適應性小腦模型控制器可以準確地控制電動運動平台之腳長長度，進而將達到平台姿態控制。

In this study, we have designed an electric motion platform, and developed an adaptive cerebellar model articulation controller to solve the nonlinear and complex dynamics of electric motion platform. Compare with oil motion platform which has existed in our laboratory, the electric motion platform has the advantages of the low noise and the low pollution (no leakage of oil problem). The electric motion platform even just need the power of 110v to drive the servo motor to turn run, and makes motion platform actions. Considering these advantages, the overseas companies which makes motion platform, is beginning to develop and produce the electric motion platform to replace the oil motion platform. Moreover, the locally and overseas be interrelated research unit as regards the needs of motion platform force in electric motion platform. In this study, two research have been discussed. One is the mechanism hardware equipment, and the other is the control algorithm design. For the mechanism hardware equipment, our study have completed the electric motion platform mechanism design､process and constructed of driving system. For the control algorithm design, since the dynamics of the electric motion platform is nonlinear and complex, the conventional control design technique is difficult to design. To solve this problem, a structure easy, fast learning property, and model-free design algorithm which is referenced as adaptive cerebellar model articulation controller system is developed. Finally, experimental results show that the adaptive cerebellar model articulation controller can achieve favorable tracking with unknown the dynamic model of the motion platform.