A Reinforcement Learning-based Decision Framework for assessing Health/Economics Dilemma in Pandemic Control

Emerging trends of AI support tools in Intelligent Healthcare (I-Health) applications have come along with recent pandemic crisis. Extreme social distancing policies, like those adopted during COVID-19 era, had in fact a burdensome impact on the worldwide economy, bringing more than ever this aspect to the attention of policy-makers. Here, we propose a Reinforcement Learning (RL) decision-maker that optimizes social distancing in an early pandemic stage, to balance social and economic costs. The epidemiological model under analysis comprises phenomena like asymptomatic infections, healthcare facilities saturation, quarantine and recovery. We compare the performance of the trained neural network with an open-loop benchmark solution obtained from direct optimization of the model. The resulting policy is remarkably alike the optimal one, with the advantage of providing a closed-loop controller function of the ongoing pandemic status, which may assist government decisions real-time.