Abstract
Critical infrastructure networks are vital for a functioning society and their failure can have widespread consequences. Decision-making for critical infrastructure resilience can suffer based on several characteristics exhibited by these networks, including (i) that there exist interdependencies with other networks, (ii) that several decision-makers represent potentially competing interests among the interdependent networks, and (iii) that information about other decision-makers' actions are uncertain and potentially unknown. To address these concerns, we propose an adaptive algorithm using machine learning to integrate predictions about other decision-makers' behavior into an interdependent network restoration planning problem considering an imperfect information sharing environment. We examined our algorithm against the optimal solution for various types, sizes, and dependencies of networks, resulting in insignificant differences. To assess the proposed algorithm's efficiency, we compared its results with a proposed heuristic method that prioritizes, and schedules components restoration based on centrality-based importance measures. The proposed algorithm provides a solution sufficiently close to the optimal solution showing the algorithm performs well in situations where the information sharing environment is incomplete.