My research has been sponsored by the National Science Foundation (NSF), Air Force Office of Science and Technology (AFOSR), Jeffress Trust Awards Program in Interdisciplinary Research, UChicago Argonne LLC, Oak Ridge Associated Universities, 4-VA, and Office of Naval Research. I am currently leading a project that studies bus fleet electrification in collaboration with school districts, transit agencies, OEM, electric utilities, local municipalities, national labs, University of Virginia, and Syracuse University. I am leading another project investigating the strengthening of the Amercian electricity infrastructure for large-scale EV adoption. I am also collaborating with researchers from George Washington University a project to explore human digital twin of neurophysiological modeling and uncertainty quantification.
My research portfolio consists of: 1) foundational and algorithmic research in the general fields of artificial intelligence/machine learning and digital twin/computer simulation; and 2) interdisciplinary applications that focuses on improving the resilience and economic efficiency of infrastructure (power grid, health care, transportation, etc.) and cyber-physical systems (cloud computing, semiconductor manufacturing, new product development, revenue management, etc.). Together with Professor Barry L. Nelson and Professor L. Jeff Hong, we developed the discrete optimization via simulation solver Industrial Strength COMPASS (ISC). ISC can be downloaded here. There are three pillars of my research: • Stochastic optimization of digital twin/computer simulation model. This is a method for the optimization of systems modeled by a digital twin/computer simulation model, or more generally, any computational models for which there is no structural information, e.g., derivative or convexity. My research focuses on innovative techniques to use adaptive stochastic search and machine learning to achieve both convergence and computational efficiency. The developed algorithms have broad applications in domains ranging from improving power grid weather resilience to hyper-parameter tuning for machine learning models. • Learning to optimize. My research focuses on optimally learning information about alternative actions/decisions to minimize loss functions defined on the outcome of the sampling process, e.g., probability of incorrect selection. This is closely related to reinforcement learning, and specifically the best arm identification problem in multi-armed bandit problem. It is also related to A/B testing. • Optimization, testing, and certification of complex systems with uncertainty. Using advanced techniques from design of experiment, machine learning, stochastic optimization, and rare event simulation, my research contributes to the optimization, testing, and certification of complex systems with uncertainty, e.g., autonomous systems, cyber-physical systems, artificial intelligence/machine learning systems, found in diverse applications such as fast recovery from supply chain disruption, production planning in semiconductor manufacturing systems, and economic optimization of cloud computing capacity. |
