| Techniques exist for assessment, modeling, and simulation of physical and cyber infrastructures, respectively; but such isolated analysis is incapable of fully capturing the interdependencies that occur when they intertwine to create a cyber-physical system (CPS). The first contribution of this doctoral research includes qualitative representation of the operation of a CPS in a single multi-agent model. Dependable operation of a CPS is contingent upon correct interpretation of data describing the state of the system. To this end, we propose agent-based semantic interpretation services that extract useful information from raw sensor data. We utilize the summary schemas model to reconcile differences in data resolution, syntax, and semantics; and to facilitate imprecise query of databases that maintain historical information, including failure mitigation techniques.;Another contribution of the research is in developing ontologies that enable automated reasoning in the classification and mitigation of failures in CPS operation. As a measure of dependability, we quantify the effectiveness of our proposed ontology- based approach in identifying correct mitigation techniques. Our methodology and models are applicable to a broad range of CPSs; however, they are described in the context of intelligent water distribution networks (WDNs), which are cyber-physical critical infrastructure systems responsible for reliable delivery of potable water. We illustrate the use of game theory in agent-based decision support for allocation of water. As a precursor to empirical validation with field data, we developed an integrated cyber-physical WDN simulator using EPANET and MATLAB, and illustrate the use of this simulator in validating our agent-based model and ontology-based approach to automated mitigation of failure. |
