- This EArly-concept Grant for Exploratory Research (EAGER) project will develop new mathematical foundations and computer-based learning theories for generating a wide range of simulated and fully-synthetic datasets that model interdependent communications and energy infrastructures in urban settings. These enhanced datasets and associated data building tools will provide a large-scale test data related to interdependent critical infrastructures (ICIs). New simulated and synthetic data generation tools will enable increasing the resiliency and flexibility of ICIs, improving their security during extreme weather conditions and other threats. This project will involve students from diverse backgrounds in engineering, computer science and psychology, who will be trained on pertinent research approaches related to the challenges of simulated and synthetic data modeling. The education plan includes a new data-centric course called Methods for Creating Simulated and Synthetic Data, as well as a large-scale involvement of graduate and undergraduate students in big data and smart community research. Broad dissemination is ensured by enabling an open-access repository of datasets created from the results of the funded research, as well as any program codes or related tools used to generate and analyze such data. The open-access testbed is capable of supporting both the research needs of the host institution as well as the requirement of non-proprietary multi-domain open datasets by other users.This project will develop a scientific basis for the generation of simulated and synthetic data on ICIs, such as communication and energy. The objective is to develop models that can accurately reconstruct, simulate, and evaluate a robust theoretical framework of ICI function by leveraging available real-world datasets. This research will lead to several innovations: 1) An advanced Transfer Learning technique that generates simulated data on ICIs using available real-world information, leading to improved characterization of how interdependencies can form or disappear over time; 2) a Hierarchical Bayesian method-based technique for the creation of synthetic data that enables ICIs to optimally manage their shared resources in response to failures from day-to-day operations, natural disasters, or malicious attacks; 3) a Long-/Short-term Memory-based deep learning method for predicting simulated data on human-in-the-loop cognitive modeling of human behavior and the effects of their decision-making in response to unexpected incidents and events involving urban ICIs; and 4) a quality-feedback loop verification and data management approach to fine-tune the simulated and synthetic data by comparing it against available real-world data over a realistic network with a large-scale simulator that integrates ICIs over an urban setting.
- September 1, 2017 - August 31, 2021
sponsor award ID
local award ID
- Sarwat, Arif Principal Investigator