Microsoft AI Framework Optimizes Pandemic Lockdown Policies
Researchers from Microsoft and partner institutions develop an AI system that automatically learns optimal lockdown and social distancing policies based on disease parameters. The framework outperforms traditional modeling tools by simulating real-world population movement and infection rates.
Researchers from Microsoft, the Indian Institute of Technology, and TCS Research develop an AI framework that helps cities and regions make policy decisions about lockdowns and physical distancing during pandemics like COVID-19. The system automatically learns optimal policies based on specific disease parameters, including infectiousness, gestation period, death probability, population density, and movement propensity, which makes it superior to traditional modeling tools.
The team creates a graph network with 100 nodes representing cities or regions and 1,000 individuals to simulate real-world conditions. The strength of connections between these nodes is directly proportional to the population and inversely proportional to the distance between them. Researchers then model COVID-19 parameters, assuming a 5- to 10-day incubation period, a 7- to 14-day infected period, an 80% likelihood of visible symptoms, a 2% death rate, and a 100% transmission probability for contact between infected and susceptible persons.
The simulation establishes specific travel rules, allowing asymptomatic and exposed individuals to move freely between open nodes while restricting symptomatic individuals. When a node undergoes a lockdown, all travel to and from that location is completely blocked. The system also accounts for human behavior by factoring in that a small number of symptomatic people break quarantine and circulate within their nodes, providing a highly realistic model for government use.