Identifying Critical Sub-systems in the Simulation of Cyber-physical Systems

The great challenge in designing dependableCyber-Physical Systems (CPS) is to structure them as robust and decoupled sub-systems, where individual sub-system failures will only have a limited impact on the overall CPS\´performance. In this paper, we discuss a CPS performing as a building emergency evacuation assistant. The CPS that weconsider in this paper, relies on a network of wired hazardsensors, computers and RF devices to provide the evacuees with personalized evacuation information, and it is used to model the effect of a fire, which not only poses a threat to the building\´s occupants but also damages the CPS\´ sensors and other systems. This paper introduces a graph theory-based generic method to identify and rank "critical" sensors, i.e. those whose information is most valuable. We validate the proposed approach through a simulation in which the critical sensors are"hardened", thereby extending their lifetime under fire. Our simulations show how the proposed approach which selects a few critical sensors that are to be hardened, can improve the outcome of the evacuation.