The EMERGE project (ExaEpi for Elucidating Multiscale Ecosystem Complexities for Robust, Generalized Epidemiology) is building a next-generation agent-based model (ABM) that captures how diseases spread both directly between people and through environmental pathways such as air, insects, and water. An ABM is like a computerized “mini-world” made up of many individual characters—called agents—that each follow a small set of rules. Every agent might represent a person, an animal, a car, or even a cell. The simulation lets these agents move around, bump into one another, make decisions, and change over time. By watching millions of simple agents interact with themselves and their environment, researchers can see how large-scale patterns—such as a traffic jam forming, a disease spreading, or an economy growing— emerge naturally. Thus an ABM starts with simple rules for individuals and ends up revealing complex behavior for the whole system. ABMs have been used to study a wide variety of phenomena spanning several different communicable human diseases, to business models involving production, selling and consuming, to even some nascent work on cancer and diabetes. However, their use for forecasting and control has been limited due to difficulties in calibrating them to the multitude of data streams available during an outbreak and quantifying the uncertainties of the model. 

Over the coming year, large ensembles of simulations using the exascale-capable ABM code ExaEpi will be carried out on DOE ASCR compute facilities to pinpoint the most influential variables for the COVID-19 outbreak. The refined set will train fast AI surrogate models that can be quickly calibrated to data such as hospitalizations, deaths, testing results, wastewater, etc. Applying the AI technique called reinforcement learning, EMERGE will create a foundation for a real-time decision system that tests a variety of intervention policies and their uncertainties. 

By fusing massive computing power, diverse data streams, and AI, EMERGE is expected to deliver rapid forecasts with quantified uncertainties, giving public-health officials a clearer picture of how policy choices may play out days to years in the future for a variety of known and potential disease outbreaks. The work advances DOE’s mission by pushing the limits of exascale computing, data integration, and AI, while also strengthening national preparedness for emerging biological threats.