Nonpharmaceutical Stochastic Optimal Control Strategies to Mitigate the COVID-19 Spread

Abstract

This article proposes a stochastic nonlinear model predictive controller to support policymakers in determining robust optimal nonpharmaceutical strategies to tackle the COVID-19 pandemic waves. First, a time-varying SIRCQTHE epidemiological model is defined to get predictions on the pandemic dynamics. A stochastic model predictive control problem is then formulated to select the necessary control actions (i.e., restrictions on the mobility for different socioeconomic categories) to minimize the socioeconomic costs. In particular, considering the uncertainty characterizing this decision-making process, we ensure that the capacity of the healthcare system is not violated in accordance with a chance constraint approach. The effectiveness of the presented method in properly supporting the definition of diversified nonpharmaceutical strategies for tackling the COVID-19 spread is tested on the network of Italian regions using real data. The proposed approach can be easily extended to cope with other countries’ characteristics and different levels of the spatial scale. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —This article is motivated by the emerging need for developing effective methods to support policymakers in mitigating the effects of the COVID-19 pandemic. The proposed feedback control strategy—combining a multiregion epidemiological model with a nonlinear stochastic model predictive control approach—allows the robust identification of the most effective restrictive measures considering the corresponding effects on the healthcare and socioeconomic systems. The proposed framework is a general and flexible method that can be applied to various real scenarios, leveraging mobility data, available from the Google mobility service, to recognize patterns and predict future behaviors of individuals.