Quantifying the Resiliency of Replicated Networked Control Systems to Transient Faults

Networked control systems (NCS) [5]—where sensors, controllers, and actuators belonging to different control loops are connected through a shared network—are highly susceptible to both internal and external sources of electromagnetic interference (EMI), e.g., engine movements, TV towers, etc. [7]. System engineers thus use active replication (or static redundancy) for ensuring that safety-critical NCSs are failoperational [4, 6, 8]. Passive replication techniques, such as the use of hot/cold standbys, are insufficient in this regard because of the time-sensitive nature of NCSs. However, coming up with a good scheme for static redundancy is a challenging problem for the following two reasons. First, any scheme should satisfy both dependability requirements and size, weight, power, and cost constraints of the platform. Second, it must take into account the inherent robustness of controllers for accurate reliability modeling. Our goal is to develop a reliability analysis that quantifies the resiliency of safety-critical, CAN-based NCSs with active replication towards EMI-induced transient failures. The analysis will provide system engineers with a method to evaluate their design choices w.r.t. the overall system reliability, and particularly evaluate any reliability bottlenecks in the designs.