Navigating Safe Campus Operations during Epidemics with Reinforcement Learning
Keywords: Q-Learning, Epidemic Control, Epidemic Simulation, Public Health Policies
Abstract: Epidemic modeling, which includes both deterministic and stochastic methods, has been central to understanding infectious disease dynamics and guiding public health decisions. While a significant portion of machine learning research in this domain focuses on predictions and trends of the disease, this study takes a prescriptive approach. This work introduces SafeCampus \footnote{Github repository to be made public}, a tool that simulates infection spread and facilitates the exploration of various RL algorithms in response to epidemic challenges. The focus is in using reinforcement learning (RL) to develop occupancy strategies that could balance minimizing infections with maximizing in-person interactions in educational settings. SafeCampus incorporates a custom RL environment, leveraging a stochastic epidemic model, to realistically represent university campus dynamics during epidemics. We evaluate a Q-learning algorithm in this context for a discretized state space to yield a sensible policy matrix, which prescribes decisions about the level of occupancy suitable for different epidemiological phases.
Submission Number: 14
Loading