網路控制系統之時間延遲控制與訊息估測之分析

Abstract

網路控制系統(NCS)改進了傳統的點對點連接方式，提供了安裝、使用以及擴展方面的便利性。許多研究著重在網路訊號延遲和資料遺失這兩個主題上。本論文提出了一個針對PI網路控制器、以系統頻寬為目標的參數自動調整方法。經由控制器參數的重新調整，可以在控制帶有變動因素的平臺時達到特定的系統頻寬要求，並且能夠降低因網路延遲時間變化產生的系統穩定性劣化程度。因為PI控制器在網路控制上有其極限，本論文提出了一個採用H-infinity演算法的網路控制器設計流程，並且演示於一經系統辨別而得的伺服控制系統模型上。在閉迴路的控制架構之外，額外加上了一個前饋式補償器，以符合對於系統頻寬及延遲容忍度雙方面的要求。在資料經網路傳送後遺失的情況發生時，本論文提出了以估測的方式重建訊息使系統得以維持正常運作，追跡效能也不至於過度劣化。在控制器下命令至驅動器的路徑上，使用了對過往命令作多項式近似的MFLSF訊息估測器；在感測器回傳系統動態資料至控制器的路徑上，則使用了建立在系統模型上的MBLSF訊息估測器。在以馬可夫鏈為網路模型的前提之下，探討了這兩類型的訊息估測器，並在不同網路情境下討論其可用性與追跡效能。

Networked control systems (NCSs) offer advantages over the traditional point-to-point connections and has been studied mainly in the network-induced delay and the package dropout. Many theoretical approaches has been proposed to handle this two problems with various algorithms. In this thesis, a general PI controller auto-tuning method based on the optimization process is proposed. The specified system bandwidth can be achieved after the tuning process without knowing the exact plant model applied to NCS. To handle the network delay, actual procedures for H-infinity controller design for NCSs is presented under the mixed synthesis framework. The control structure with a forward compensator is used to meet specifications in both delay tolerance and bandwidth. For compensating network dropout in controller-to-actuator and sensor-to-controller paths, both the modelfree least square fit (MFLSF) estimator and the model-based least square fit estimator (MBLSF) are used to improve the tracking performance, respectively, and stability criteria of both MFLSF and MBLSF estimators are provided with the Markov chain model. By switching MFLSF estimators according to the current identified network status, total tracking performances can be thus significantly improved.