REINFORCEMENT LEARNING-BASED POST DISASTER RESOURCE ALLOCATION FOR ENHANCED INFRASTRUCTURE RESILIENCE

Author

Sandeep Acharya, Southern Illinois University CarbondaleFollow

Date of Award

8-1-2024

Degree Name

Master of Science

Department

Civil Engineering

First Advisor

Sen, Debarshi

Abstract

Natural calamities such as floods, earthquakes, fire are increasing, and infrastructure networks are exposed to such risks during their lifetime. Modern infrastructures are interconnected with each other and depend on one another to fulfill their intended purpose, making the analysis of the network harder. This necessitates the enhancement of the resilience of the infrastructure during recovery. The recovery strategy adopted without the interdependencies is not optimal. In this study, the framework to allocate the resources (human resources and capital) effectively to improve the overall infrastructure system resilience considering the interdependencies among and within the infrastructure facilities is proposed. Typically, the resource allocation is treated as a high-dimensional, multi-objective optimization problem. In this study, this task was formulated as sequential decision-making problem. By using Agent-Based modelling to simulate the effects of infrastructure interdependencies, and deep Reinforcement Learning to solve the sequential decision-making problem, the proposed framework can find the resource allocation strategies enhancing the overall infrastructure systems resilience. Particularly, the aim of this study is to allocate the limited number of available repair crews to the damaged facilities within budget constraint. The results illustrate the efficacy of the proposed framework in simulating interdependency and finding the allocation strategies for post-disaster scenarios.