Solving the static INRC-II nurse rostering problem by simulated annealing based on large neighborhoods

Abstract

This paper proposes a local search method based on a large neighborhood to solve the static version of the problem defined for the Second International Nurse Rostering Competition (INRC-II). The search method, driven by a simulated annealing metaheuristic, uses a combination of neighborhoods that either change the assignments of a nurse or swap the assignments of two compatible nurses, for multiple consecutive days. Computational results on the set of competition instances show that our method has been able to improve on all previous approaches on some datasets, and to get close to the best ones in others. Best solutions, along with the datasets and the validation tool, are made available for future comparison.