Nurse Rostering and Integer Programming Revisited

Design and development of robust and reliable scheduling algorithms has been an active research area in Computer Science and Artificial Intelligence. Given that the general problem is computationally intractable, many heuristic-based techniques have been developed, while other approaches have used optimising techniques for specific and limited problem domains. In this paper, we consider a large, real world scheduling problem of nurse rostering at the Gold Coast Hospital, and we propose an optimising integer programming-based approach to solving the problem. The study extends previous work in the area by looking at a more complex problem domain and by introducing a mathematical model that is capable of capturing the important details of the domain without becoming unrealistically large. We also describe a problem decomposition heuristic to effectively manage the computational resources. Schedule quality and staff allocation quality measures are introduced to evaluate the automated nurse schedules. Finally, we conclude that an integer programming-based approach to nurse rostering is feasible for realistic problem sizes and is sufficiently flexible to handle overconstrained problems with competing goals.