On the performance optimization of reliable networked control systems

Networked Control Systems (NCSs) consist of sensors, controllers and actuators which are all interconnected via a network fabric. Such an architecture has proven advantageous for industrial communication networks compared to traditional approaches. Lately, Wireless NCSs (WNCSs) have also been rising in popularity due to their ease of installation and maintenance compared to equivalent wired NCSs. However, a significant drawback of WNCSs is the experienced wireless interference which arises from the shared nature of the wireless medium. In any NCS, the failure of a single component can cause complete control system failure in the absence of fault-tolerance. As a result of the ensuing system downtime due to such failures, large production losses could potentially occur. Thus, fault-tolerance is now becoming a crucial aspect of the design and evaluation of NCSs. Fault-tolerance can be implemented at various levels of an NCS in order to improve system reliability: either at the node level and/or at the network fabric level. Nevertheless, the incorporation of fault-tolerance in NCSs involves additional overhead traffic which can have a noticeable impact on the overall system performance. This overhead traffic may cause the real-time NCS to miss crucial control deadlines. Therefore, minimizing the amount of traffic overhead necessary for the implementation of fault-tolerance is desired. This research is focused on the design and performance optimization of reliable fault-tolerant NCSs and WNCSs. First, a fault-tolerant WNCS is proposed based on unmodified IEEE 802.11b implementing 1-out-of-3 controller level fault-tolerance utilizing a wired backbone. The interference tolerance of the system was quantified and certain performance optimizations were investigated in order to improve the overall system’s interference resilience. Moreover, an additional fault-tolerant WNCS with a reliable wireless backbone is proposed. The proposed WNCS is based on unmodified IEEE 802.11g and implements 1-out-of-2 controller level fault-tolerance in addition to network fabric level fault-tolerance on the critical wireless backbone link using the Parallel Redundant Protocol (PRP). Second, a network fabric fault-tolerance methodology is investigated for wired Ethernet NCSs utilizing the Rapid Spanning Tree Protocol (RSTP). A performance optimization is proposed which halves the amount of traffic necessary for the implementation of fault-tolerance while guaranteeing system resilience to any individual network fabric failure. Furthermore, a reliability modeling methodology is developed for the proposed model. A case study is subsequently presented to compare reliability of different system architectures using typical industrial parameters. Finally, an expanded two cell model is developed which not only provides the same degree of network fabric level fault-tolerance but also controller level fault-tolerance