網路控制系統之時間延遲補償設計

Abstract

近年來隨著網路的興起，即時的網路控制成為一種趨勢，在工業上的應用也越來越多，其可以很便利達成系統性的維護。然而，將控制系統網路化之後，也帶來了幾個缺點，例如：在共享的有限網路資源，隨著使用者數目的增減而造成變化極大的時間延遲，此問題輕則明顯降低系統效能，重則使整個系統產生不穩定的情形。 根據網路協定、節點數和軟硬體條件，網路的時間延遲特性可能是固定或是有界的，甚至是隨機和不可預測的。因此，針對處理實際網路的重大的時間延遲變化，本論文提出兩種網路控制系統(NCS)之時間延遲補償的方法。第一個方法是發展即時的時間延遲估測，透過量測實際網路環境中兩個節點封包往返的時間(RTT)，可估測網路控制系統的時間延遲，並應用於三個方面：（1）發展適應性史密斯預估控制，可針對重大的時間延遲變化作處理；（2）強健性的網路控制系統設計，可對付具有局部時間延遲變化與外部干擾的NCS；及（3）多重取樣週期的設計，可針對無線網路的壅塞問題作解決。上述所提之方法均已成功地實現於一交流伺服馬達的遠端控制系統。 此外，第二個解決時間延遲的方法，是提出時間延遲完全補償策略(PDC)，能有效處理網路所引起時間延遲的影響，既不需要系統的模型也不用已知時間延遲的資訊。網路控制系統能等效為原來的閉迴路系統串接一單純的時間延遲。因此，當引用PDC在設計網路控制系統時，不需考慮對網路時間延遲的影響，只需將所設計的控制系統直接實現於網路上即可。最後，實驗結果透過十五公里的Internet網路連線，進一步證明SISO和MIMO控制系統，都可經由所提出的PDC直接施行網路化，在實際網路環境保持其系統閉迴路的特性。

Real-time network control applications have increasingly gained attentions due to the rapid development of data communication network technologies. Network systems can be conveniently and systematically maintained in industrial applications. The networked control system (NCS), which simply interconnects all sensors, actuators, and controllers through the network, is promising in the future development of industrial technologies with low integration cost. However, NCS also leads to unavoidable problems in time delays that seriously degrade control performance and stability. The characteristics of network-induced delays may be in constant, bounded, random, or unpredictable natures depending on the network protocols, nodes, software, and hardware. In this study, there are two approaches proposed for NCS design under significantly varied time delays. In the first approach, on-line estimation of the delay time is developed by processing the on-line measurement of the round-trip time (RTT) between two nodes in real network environments. Three related controllers are thus developed: (1) the adaptive Smith predictor control scheme for significantly varied time delay, (2) the robust NCS design for bounded variation of time delay and disturbance, and (3) the multi-rate design under the condition of wireless network congestion. The second approach is proposed as the model-free perfect delay compensation (PDC) scheme. This scheme effectively deals with network-induced delays requiring neither the delay time information nor the plant model. NCS with PDC is thus simply equivalent to the original closed-loop system with an additional pure time delay and it is designed without concerning the network. Therefore, the well-designed controller can be directly implemented on a network and its stability can be guaranteed without being affected by the varied time delay. The proposed approaches have been successfully applied to remote control systems under significantly time-varying delay to control an AC servo motor. Provided experimental results have further proven that both SISO and MIMO systems can be directly implemented in networking systems by including the proposed PDC to maintain its original feedback-loop characteristics.