Pieter Smet

Modelling and evaluation issues in nurse rostering

In practice nurse rostering problems are often too complex to be expressed through available academic models. Such models are not rich enough to represent the variegated nature of real world scenarios, and therefore have no practical relevance. This article focuses on two particular modelling issues that require careful consideration in making academic nurse rostering approaches re-usable in a real world environment. First: introducing several complex problem characteristics, resulting in a rich, generic model. A detailed description is provided for researchers interested in using this new model. We also present a novel benchmark dataset based on this rich model. Second: the consideration of a consistent evaluation procedure that corresponds to realistic quality measurement. These contributions will enable faster implementation of academic nurse rostering achievements in real hospital environments. A suite of hyper-heuristics is presented. These are capable of solving these rich personnel rostering problems using the presented evaluation procedures. Their performance is compared to that of another meta-heuristic.

The Multi-Mode Resource-Constrained Multi-Project Scheduling Problem

Scheduling projects is a difficult and time consuming process, and has far-reaching implications for any organization’s operations. By generalizing various aspects of project scheduling, decision makers are enabled to capture reality and act accordingly. In the context of the MISTA 2013 conference, the first MISTA challenge, organized by the authors, introduced such a general problem model: the Multi-Mode Resource-Constrained Multi-Project Scheduling Problem (MRCMPSP). The present paper reports on the competition and provides a discussion on its results. Furthermore, it provides an analysis of the submitted algorithms, and a study of their common elements. By making all benchmark datasets and results publicly available, further research on the MRCMPSP is stimulated.

An analysis of generalised heuristics for vehicle routing and personnel rostering problems

The present study investigates the performance of heuristics while solving problems with routing and rostering characteristics. The target problems include scheduling and routing home care, security and maintenance personnel. In analysing the behaviour of the heuristics and determining the requirements for solving the aforementioned problems, the winning hyper-heuristic from the first International Cross-domain Heuristic Search Challenge (CHeSC 2011) is employed. The completely new application of a hyper-heuristic as an analysis tool offers promising perspectives for supporting dedicated heuristic development. The experimental results reveal that different low-level heuristics perform better on different problems and that their performance varies during a search process. The following characteristics affect the performance of the heuristics: the planning horizon, the number of activities and lastly the number of resources. The body of this paper details both these characteristics and also discusses the required features for embedding in an algorithm to solve problems particularly with a vehicle routing component.

Nurse Rostering: a Complex Example of Personnel Scheduling with Perspectives

Nurse rostering is an attractive research domain due to its societal relevance, while academics are intrigued by its combinatorial complexity. Descriptions of nurse rostering problems vary largely across the literature, which makes it almost impossible to track down scientific advances of models and corresponding approaches. The present chapter introduces a mathematical formulation of a generic nurse rostering model. It provides common elements present in most nurse rostering research as well as important hospital constraints that are usually omitted from academic models. The new mathematical model satisfies all the basic requirements for future nurse rostering research and practical developments. Finally, the importance of public datasets is discussed, together with the characteristics of the various benchmark instances and research results obtained working on these instances.

Heuristic decomposition approaches for an integrated task scheduling and personnel rostering problem

Real world combinatorial optimisation problems do not often reduce to neatly delineated theoretical problems. Rather, they combine characteristics of various subproblems which then appear to be strongly intertwined. The present contribution introduces a challenging integration of task and personnel scheduling in which both tasks and shifts must be assigned to a set of multi-skilled employees. Three constructive heuristics, based on column generation and other decomposition schemes, are presented, as well as a very large-scale neighbourhood search algorithm to further decrease the schedules cost. The performance of these algorithms is evaluated on a large set of diverse instances. Computational results illustrate the effectiveness of the proposed approaches, and provide insight into their behaviour. The initial benchmarks are published so as to encourage further research. HighlightsAn integrated task scheduling and personnel rostering problem is introduced.Two integer programming formulations are presented and experimentally evaluated.Stabilised column generation is used in a constructive heuristic.Experiments compare constructive heuristics and very large-scale neighbourhoods.

Polynomially solvable personnel rostering problems

Personnel rostering is a personnel scheduling problem in which shifts are assigned to employees, subject to complex organisational and contractual time-related constraints. Academic advances in this domain mainly focus on solving specific variants of this problem using intricate exact or (meta)heuristic algorithms, while little attention has been devoted to studying the underlying structure of the problems. The general assumption is that these problems, even in their most simplified form, are NP-hard. However, such claims are rarely supported with a proof for the problem under study. The present paper refutes this assumption by presenting minimum cost network flow formulations for several personnel rostering problems. Additionally, these problems are situated among the existing academic literature to obtain insights into what makes personnel rostering hard.

Facilitating the transition from manual to automated nurse rostering

After several decades of academic research in the field of automated nurse rostering, few results find their way to practice. Often, the configuration of a software system for automated rostering presents a task considered too time-consuming and difficult. The present article introduces a methodology for automating part of the costly and unintuitive configuration process by automatically determining the relative importance of soft constraints based on historical data. Naturally, this automated approach can only be reliable in the absence of transient effects and drastic changes. The approach is evaluated on retrospective and prospective case studies, and is validated by health-care practitioners partaking in an experimental study. The results show that, given relevant historical data, the presented approach simplifies the transition from manual to automated rostering, thus bringing academic research on nurse rostering closer to its practical application.

Local and global constraint consistency in personnel rostering

Personnel rostering is a challenging combinatorial optimization problem in which shifts are assigned to employees over a scheduling period while subject to organizational, legislative, and personal constraints. Academic models for personnel rostering typically abstractly conceptualize complex real-world problem characteristics. Often only one isolated scheduling period is considered, contradicting common practice where personnel rostering inherently spans multiple dependent periods. The state of the art offers no systematic approach to address this modeling challenge, and consequently, few models capture the requirements imposed by practice. The present paper introduces the concepts of local and global consistency in constraint evaluation processes and proposes a general methodology to address these challenges in integer programming approaches. The impact of inconsistent constraint evaluation is analyzed in a case study concerning rostering nurses in a hospital ward, of which the data have been made publicly available. The results demonstrate that the proposed methodology approximates the optimal solution.