Interdependent networks are an integral part of today’s modernized societies in which the performance of some components in one network depends on the services provided by components of other networks. Consequently, failure of a node or component in one network may cause failure in another network. Optimal operation of interdependent networks is challenging due to their coupled nature, overlapping decisionmaking variables, decentralized decision-making context, limited information sharing among networks, and potentially conflicting utility functions. These networks, including critical infrastructures (CIs), are pivotal for societal well-being, ensuring continuous access to essential services such as transportation, water, gas, and electricity. Implementing robust preparedness, response, and recovery policies for these networks significantly enhances community resilience. In this dissertation, we introduce a novel methodology, referred to as the coupled Reinforcement Learning (RL) mechanism, that leverages the strengths of RL to enhance the computational capability of optimization techniques and generate a set of cooperative restoration policies for a system of interdependent CIs. The proposed approach: (1) bridges the gap between integrative and distinct decision-making, enabling coordinated restoration planning for a set of interdependent CIs within a decentralized decision-making context with limited information sharing, (2) is capable of handling post-disaster uncertainties (e.g., uncertainty in recovery times of disrupted components), (3) generates adaptive solutions that cope with post-disaster dynamics (e.g., varying numbers of recovery teams), and (4) is flexible enough to handle several restoration decisions (e.g., restoration scheduling and resource allocation) simultaneously. Collectively, this dissertation advances our understanding of cooperative decision-making tailored for resilient interdependent networks, carrying profound implications for post-disaster CI restoration efforts in communities. By bridging the gap between integrative and distinct decision-making, our mechanism offers a robust framework for fostering interaction among various networks (e.g., CI operators) and resilience in the face of complex disaster and failure scenarios in interdependent networks. To evaluate the decision-making effectiveness of our proposed methodology, we focus on the concurrent restoration of road and power CIs in Sioux Falls, South Dakota, following several tornado-induced disruption scenarios. The findings of our research underscore the efficacy of the proposed mechanism in enhancing post-disaster restoration operations, which translates to more resiliency. Our numerical results highlight the potential of the proposed approach to mitigate the impact of disasters on CI networks and enhance societal resilience.

Namrata Saha is currently a Ph.D. candidate and serves as a Graduate Assistant at the Sustainability, Optimization, and Learning for InterDependent networks laboratory (solid lab) in Knight Foundation School of Computing and Information Sciences and the Smart Decision-Making for Network-Centric Systems Laboratory in the Enterprise and Logistics Engineering program at Florida International University. Prior to starting her Ph.D., she obtained a B.Sc. degree in Computer Science and Engineering from the Military Institute of Science and Technology, Dhaka, Bangladesh in 2018. Under the supervision of Dr. M. Hadi Amini and Dr. Shabnam Rezapour, her research focuses on developing novel decision/policy-making methodologies that integrate the strengths of Reinforcement Learning (RL) and optimization algorithms to enhance the resilience of interdependent networks against potential disruptions. She tested and applied her developed approach to generate coordinated restoration policies for interdependent critical infrastructures in disaster-prone communities. She received Doctoral Assistantship from the FIU Preeminent Institute for Resilient and Sustainable Coastal Infrastructure (InteRaCt).