CoV-ABM: A stochastic discrete-event agent-based framework to simulate spatiotemporal dynamics of COVID-19

Abstract

The paper develops a stochastic Agent-Based Model (ABM) mimicking the spread of infectious diseases in geographical domains. The model is designed to simulate the spatiotemporal spread of SARS-CoV2 disease, known as COVID-19. Our SARS-CoV2-based ABM framework (CoV-ABM) simulates the spread at any geographical scale, ranging from a village to a country and considers unique characteristics of SARS-CoV2 viruses such as its persistence in the environment. Therefore, unlike other simulators, CoV-ABM computes the density of active viruses inside each location space to get the virus transmission probability for each agent. It also uses the local census and health data to create health and risk factor profiles for each individual. The proposed model relies on a flexible timestamp scale to optimize the computational speed and the level of detail. In our framework each agent represents a person interacting with the surrounding space and other adjacent agents inside the same space. Moreover, families stochastic daily tasks are formulated to get tracked by the corresponding family members. The model also formulates the possibility of meetings for each subset of friendships and relatives. The main aim of the proposed framework is threefold: to illustrate the dynamics of SARS-CoV diseases, to identify places which have a higher probability to become infection hubs and to provide a decision-support system to design efficient interventions in order to fight against pandemics. The framework employs SEIHRD dynamics of viral diseases with different intervention scenarios. The paper simulates the spread of COVID-19 in the State of Delaware, United States, with near one million stochastic agents. The results achieved over a period of 15 weeks with a timestamp of 1 hour show which places become the hubs of infection. The paper also illustrates how hospitals get overwhelmed as the outbreak reaches its pick.