CAREER: Simulation Optimization Reimagined: Coupling Exploratory Simulation Analysis and Optimization for Holistic Decision Making

About This Grant

This Faculty Early Career Development Program (CAREER) grant will advance the national prosperity and economic welfare by enhancing the analytical capabilities of organizations in sectors such as healthcare, finance, construction, and national defense that leverage stochastic computer simulation models to make critical decisions in the face of uncertainty. This award supports a fundamental reinvention of how such models are paired with optimization methods to inform decision makers of risks and tradeoffs in stochastic system performance. This research will make simulation optimization approaches more systematic, productive, and aligned with user needs and facilitate more holistic decision making than conventional approaches. Close collaboration with industry partners will ensure the methods created are intuitive, informative, and practicable. The educational component of the project will create high school outreach activities and teaching modules that explore analysis techniques for simulation data and improve programming proficiency and statistical literacy. This project will also produce software, including open-source implementations of the methods, a prototype of an interactive dashboard, add-ins for commercial simulation software, and versions that are compatible with an open-source simulation optimization testbed used by researchers and educators. The research is motivated by shortcomings of existing simulation optimization (SO) approaches, which generally require decision makers to specify summary performance measures to serve as objectives or constraints in an optimization problem. By beginning with a narrow problem formulation, SO practitioners often fail to think about their simulation model in the broadest stochastic sense. This research shifts the initial focus of SO from summary performance measures to distributions of performance measures, exposing the user to inherent risks and tradeoffs. The research invents a transformative framework that couples exploratory simulation analysis with powerful optimization technologies to facilitate more holistic decision making. This project will create new search methods for discovering solutions with differing output distributions, incorporate user input in pursuit of optimization goals, exploit parallel computing resources to accelerate the search and optimization processes, and extend the framework to handle simulation trace data. The research will require the invention of new methods for dynamically clustering multivariate distributions and stochastic processes and metamodeling simulation outputs and output distributions that will be rigorously analyzed from both a theoretical and computational perspective. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.