網路控制系統之智慧型資料遺失補償器設計

Abstract

網路化控制在工業上已經成為一種趨勢。然而，網路頻寬的限制導致時間的延遲和資料的遺失等，輕則明顯降低系統效能，重則使整個系統產生不穩定。本論文提出兩種即時網路控制系統(NCS)之資料遺失補償的方法來改善系統的效能。第一個方法是針對不同的資料遺失率做補償，提出泰勒訊息估測器：第二個方法是針對不同分佈的遺失資料，提出即時轉移機率估測器和智慧型訊息補償器。即時轉移機率估測器利用即時短時間內訊息遺失的機率快速地計算出轉移機率，並且成功且有效率的監控資料遺失的分佈情形。本論文分別從追跡效能和穩定度上分析出不同階數的最小方差估測器可適用的轉移機率範圍，以即時估測出的轉移機率來切換不同的補償器，進而建構出智慧型訊息補償器(IME)來有效的補償不同分佈的資料遺失效應。但是當系統出網路時間延遲時，單獨IME是無法處理的，本論文整合了IME和perfect delay compensator (PDC)達成了同時處理資料遺失和網路時間延遲的效應。經模擬及實驗證明本論文所提出不需模式之方法，皆能有效且簡單應用在實際運動控制系統上，達成網路時間延遲及訊息遺失的補償效果，達成系統的精密度。

The time delay and the dropout in network control systems (NCS) are unavoidable mainly because of their limited transmission bandwidth. In this study, the Taylor estimator scheme is developed to effectively improve tracking accuracy of motion NCS under a higher data dropout rate in uniform distributions. Furthermore, as the data dropout in real applications is in stochastic nature with time-varying properties, the missing messages occur consecutively to cause the message compensator based on uniform distribution of data dropout to be invalid. A real-time transition probability estimator is thus proposed to monitor the data dropout distribution in real time. With the least square algorithm, estimators with real-time orders according to different network conditions are then studied. In this dissertation, the proposed model-free intelligent message estimator (IME) is developed to compensate for different data dropout distributions in a switching mechanism based on the estimated transition probability and stability analysis. However, when both the time delay and the data dropout are simultaneously induced in NCS, the IME only itself cannot provide appropriate compensation for their mutual effects. The integrated compensator combining both the IME and the perfect delay compensator (PDC), which is designed to compensate for the time delay, leads to satisfactory compensation results for both the time delay and the data dropout. Both simulation and experimental results indicate that the integration of the IME and the PDC compensator has effectively improved control performance even with varied networked time delays and data dropout distributions.