This talk considers a dynamic Emergency Medical Services network design problem and introduces two novel two-stage stochastic programming models that account for uncertainty about emergency demand. Similarly to some recent work on the emergency demand coverage model, we consider both a constraint on the probability of covering the realized emergency demand and minimize the expected cost of doing so, yet unlike other formulations, our models account for the dynamics throughout a full day of operations and allow the EMS system managers to control the degradation of coverage under the more severe scenarios. In order to do so, we present both a two-stage chance-constrained stochastic programming formulation and a variant of this model that employs probabilistic envelope constraints. These give rise to large mixed-integer programs, which can be tackled directly or using a conservative approximation scheme in a Branch-and-Benders-Cut framework. Finally, a practical study is conducted using historical data from Northern Ireland and sheds some light on optimal EMS network configuration and necessary trade-offs between coverage level and expected cost. These insights are confirmed through an out-of-sample analysis.
Coffee and biscuits will be offered at the beginning of the seminar.
Welcome to everyone!