Parallelizing constraint solvers for hard RCPSP instances

The Resource-Constrained Project Scheduling Problem (RCPSP) is a well-known scheduling problem aimed at minimizing the makespan of a project subject to temporal and resource constraints. In this paper we show that hard RCPSPs can be efficiently tackled by a portfolio approach that combines the strengths of different constraint solvers Our approach seeks to predict and run in parallel the best solvers for a new, unseen RCPSP instance by enabling the bound communication between them. This on-average allows to outperform the oracle solver that always chooses the best available solver for any given instance