A rollout algorithmic approach for complex parallel machine scheduling in healthcare operations

Healthcare Operations Management calls to tackle different scheduling problems. Often the environment to consider can be modeled with a set of parallel resources to be scheduled and is characterized by one or more application-specific constraints or objectives. Rollout methodology is effortless to implement and able to easy incorporate human experience inside its research patterns to fulfill complex scheduling requirements as those of interest in healthcare applications. A drawback is represented by the high computation time often required to terminate the algorithm. We analyze a deterministic parallel machine scheduling problem showing how a rollout framework can be extended and adapted to tackle several additional specific constraints arising in healthcare operations, such as surgery block scheduling. We also describe simple methodologies that can be used to address the uncertainties in the problem by iterating the solution of deterministic models. A lower bound is used to certify the quality of the generated solutions for complex parallel machines scheduling problems. A preliminary campaign of computational experiments, shows the behavior of different algorithmic variants of this approach.