Spatiotemporal Risk-Averse Routing

A cyber-physical system is often designed as a network in which critical information is transmitted. However, network links may fail, possibly as the result of a disaster. Disasters tend to display spatiotemporal characteristics, and consequently link availabilities may vary in time. Yet, the requested connection availability of traffic must be satisfied at all times, even under disasters. In this paper, we argue that often the spatiotemporal impact of disasters can be predicted, such that suitable actions can be taken, before the disaster manifests, to ensure the availability of connections. Our main contributions are three-fold: (1) we propose a generic grid-based model to represent the risk profile of a network area and relate the risk profile to the availability of links and connections, (2) we propose a polynomial-time algorithm to identify connections that are vulnerable to an emerging disaster risk, and (3) we consider the predicted spatiotemporal disaster impact, and propose a polynomial-time algorithm based on an auxiliary graph to find the most risk-averse path under a time constraint.