Florida International University
Namrata Saha is currently pursuing a Ph.D. in computer science and serving as a Graduate Assistant at the solid lab (Sustainability, Optimization, and Learning for InterDependent networks laboratory), Knight Foundation School of Computing and Information Sciences at Florida International University. Under the supervision of Dr. M. Hadi Amini and Dr. Shabnam Rezapour, her research focuses on developing novel reinforcement learning algorithms to ensure community resilience in the face of disruptive events. Her work involves the development of a restoration policy by utilizing Deep Reinforcement Learning, Neural Network, and Optimization Algorithms. 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.
Ensuring the timely recovery of interdependent critical infrastructures is essential for enhancing the resilience of communities when faced with disruptive incidents. Interdependent critical infrastructures encompass various physical elements such as cables, transmitters, and power generators in power grids, as well as roads, highways, bridges, tunnels in road networks, and pipes and water processing facilities in water networks. These components play a crucial role in delivering essential services to a community. Given the expanding nature of communities and the growing frequency of disruptive events, critical infrastructures are susceptible to various types of disturbances. Critical infrastructures are functionally coupled. This means the functionality of components in one infrastructure depends on the services provided by other infrastructures. These interdependencies cause disruptions to cascade across communities and highly complicate their restoration processes.
The main objective of the proposed dissertation is to develop a novel modeling and solution approach that allows for the creation of a collaborative restoration plan for interconnected critical infrastructure. This problem-solving methodology operates under a decentralized decision-making structure and accounts for different levels of uncertainty. This approach involves a collection of deep reinforcement learnings (DRLs) that are interconnected through a data exchange component to facilitate information sharing among agents and enable them to make collaborative decisions. Each DRL is responsible for planning restoration activities for a specific critical infrastructure.