標題: 無線網路控制系統之訊息估測與網路流量監控
Message Estimator and Traffic Monitor for Wireless Network Control System
作者: 張哲維
Chang, Che-Wei
徐保羅
Hsu, Pau-Lo
電控工程研究所
關鍵字: 網路控制系統;資料遺失;訊息估測;網路流量監控;NCS;dropout rate;message estimator;network traffic monitor
公開日期: 2008
摘要: 網路控制系統特有的資料遺失、時間延遲對於系統效能有很大的影響。本論文提出利用訊息估測與網路流量管制的設計,可使用在以無線網路作為傳輸媒介的網路控制系統。本論文在輸出回授端採用時間觸發,具有高取樣頻率、閉迴路即時控制的優點。再透過實驗探討網路延遲時間、資料遺失率與取樣週期的關係,找出系統最佳取樣週期,作為網路遠端控制的設計。 在訊息估測方面,本研究分析不同的估測法則和網路傳輸環境對系統效能的影響,根據所使用的網路特性,選擇泰勒估測器,而在網路壅塞時,利用智慧訊息估測(Intelligent message estimator),結合泰勒估測、最小平方法與沿用前筆資料,成功解決在泰勒錯誤估測的問題,實驗結果顯示,以泰勒估測處理可降低輸出誤差約30.1%,而在網路壅塞時,以IME估測可降低輸出誤差約74.7%。 最後在網路流量監控方面,本論文針對當網路負載過大導致系統不穩定的問題,提出根據目前網路流量所成的資料遺失率來做取樣頻率的調整,實現於網路遠端監控系統,藉此降低資料遺失率約10~15% ,以提升系統控制效能與穩定度與容忍網路環境的穩健性。
The data dropout and time delay in the network control system (NCS) affect its control performance greatly. In this thesis, the message estimator and the traffic monitor designed for the wireless network control system (WiNCS) are proposed. Instead of the event-trigger NCS, the output feedback signal is sent by the time trigger and the received node implemented with a message estimator to compensate for the lost data. Different estimators in the WiNCS have been discussed and results indicate that the Taylor estimator is chosen with easy implementation on a light-traffic network communication. However, with a heavy network load, the intelligent message estimator (IME) is suitable to cope with the unstable problem in the Taylor estimator. Experiment results show that the Taylor estimator reduced the output error 30.1 % with a light network load, and the IME reduced the output error 74.7% with a heavy network load. Finally, when the network traffic is too heavy to the feedback estimator, the sampling time was adjusted according to the data dropout rate, obtained from the network traffic monitor, to further decrease the dropout rate almost 10~15% to achieve a robust NCS under different environments.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT079512506
http://hdl.handle.net/11536/41073
Appears in Collections:Thesis


Files in This Item:

  1. 250601.pdf

If it is a zip file, please download the file and unzip it, then open index.html in a browser to view the full text content.