Abstract

Robot Dance is based on a spatio-temporal model that splits the population into groups (nodes), representing the mobility as links in the network. The limit in ICU capacity, a constraint in the optimization problem, is handled in the form of probabilistic constraints. The tool can anticipate the geographical evolution of the disease and evaluate the potential impact of containment and prevention strategies. Given appropriate mobility and epidemiological input data, Robot Dance can determine different mitigation protocols that make efficient use of the ICU beds. Robot Dance pinpoints critical locations where targeted policies, of local nature, have the most impact. For the example of São Paulo in Brazil, this means that instead of putting in lockdown the whole state at once, different degrees of social distancing can be imposed in different districts, depending on the local situation of the hospitals, the severity of infections, and the exchanges that the district in question has with other nodes in the network, because of commuting. Robot Dance can also anticipate where the creation of a pool of ICU beds, to be shared by nearby districts, is most effective, considering not only the epidemiological state in the districts directly involved, but also their interaction with the whole network, through the mobility of their inhabitants. The proposed framework is highly flexible and adaptable to different available data sets and control strategies.