N次直接補償控制法於虛擬實境史都渥特平台的應用

Abstract

本文包含了實作和理論兩部分，我們提出了一個N次直接補償控制法的架構，並將此控制法應用於史都渥特平台的控制，由於N次直接補償控制法為一新的架構，我們嘗試推導出此架構的輸出誤差通式，並進一步證明N次補償的輸出響應將優於N-1次的輸出響應。在證明N次直接補償控制法具有良好的響應特性後，我們開始尋求，如何On-line預測N次直接補償控制法的補償訊號，在本文中，我們提出了兩種方法，第一種方法為Model Based，其將FIR和IIR的Model結合起來，具有FIR和IIR的優點。第二種方法為個人所提出，此方法是利用步階響應誤差和期望的輸出軌跡來預測補償訊號，其缺點為必須瞭解輸入軌跡的拉氏轉換函數，優點為精度將較第一種方法為高。為了瞭解輸入軌跡的拉氏轉換函數，我們進一步對軌跡規劃作了一番探討。另外，為了驗證N次直接補償控制法在一般系統中是否能有良好表現 ，我們嘗試將它應用於直流馬達和油壓伺服控制系統中，由模擬結果顯示，N次直接補償控制法在線性和非線性系統中具有良好的補償效果。最後，我們實際發展出一史都渥特平台的控制系統，我們對兩種常見的史都渥特平台控制架構加以探討，並提出位置、速度、加速度的開迴路架構。在此架構的前提下，我們逐步完成架構的內部細節。如位置控制器的設計。我們亦Off-line的利用N次直接補償控制法於史都渥特平台的油壓缸控制，由實驗結果發現和理論頗為耦合。

This thesis proposes an N-order feedforward compensation controller and applies it to the control of a VR-Based Stewart Platform. We have proved the convergence of the proposed feedforward control scheme;the system output error is always decreasing as the order of compensation increases. The proposed control scheme can be used in off-line applications directly to achieve high control accuracy. To apply the N-order feedforward compensation control scheme in on-line environments, we need a predictor to predict the compensation signal. Two prediction techniques are further proposed for this purpose. The first is the model-based technique, which combines the advantages of FIR and IIR models. The second prediction technique is a novel one, which predicts the compensation signal directly from the step response error of the controlled system. The second technique can obtain much higher prediction accuracy than the first one, although it needs to have the Laplace transform of the input trajectory, which can be easily obtained by proper path planning schemes. To verify the control performance of the proposed N-order feedforward compensation control scheme, we apply it to the control of DC motor and hydraulic servo systems. Simulation results show that the proposed scheme achieves very satisfactory performance either in linear and nonlinear systems. Finally, the N-order feedforward compensation control scheme is used to control a Stewart platform practically. In this real-world application, we construct three control structures for position, velocity, and acceleration control respectively. The experimental results fit our theoretic analysis very well.