Title: Advanced Models and Solution Methods for Automation of Personnel Rostering Optimisation.
Other Titles: Geavanceerde modellen en oplossingsmethoden voor de automatisering van personeel planning optimalisatie.
Authors: Bilgin, Burak; S0192350
Issue Date: 16-May-2012
Abstract: Personnel rostering is the practice of assigning daily working units, called shifts, to employees with respect to a variety of constraints, such as workforce requirements, labour legislation, union demands and individual employee preferences. The shortage of qualified personnel in the developed countries makes the personnel rostering problem one of the fundamental problems to be addressed in the private and public sector. Personnel rostering is a dynamic problem that evolves in time with the changes in the organisational structures, labour legislations and union demands. The objective of this dissertation is to present a complete solution, ready to be implemented, addressing the real world personnel rostering problem.The contribution of the present dissertation can be summarised as follows. An extended problem model is introduced to represent the complexity and the diversity of the real world problem instances. An in-depth analysis of the differences between the constraint evaluation methods of the academic literature and the real world practice is reported. The constraint evaluation methods of the real world practice are presented formally and discussed in great detail, referring to examples from the real world. An algorithmic toolbox is developed that consists of the elements from the academic literature as well as new modules that are designed to take into account the extensions made to the problem model. Three quantitative measures are proposed for assessing the properties of a problem instance. Problem instances gathered from real world sources are published as a benchmark data set. An extensive set of experiments and statistical analysis are performed on the benchmark data set and on the problem instances from academic sources. The experimental results are processed to conclude on the most suitable configuration of the algorithmic toolbox on a given problem instance. The resulting body of PhD work benefits both academia and industry, by extending the scope of the personnel rostering theory to cover the complexity and the diversity of the real world practice.
Table of Contents: 1 Introduction
2 Academic context
2.1 Nurse rostering problems
2.2 Nurse rostering benchmarks
2.3 Approaches to the Nottingham benchmarks
2.4 Problems from various sources
2.5 Real world implementation
2.6 Solution methods
2.6.1 Tabu search
2.6.2 Variable neighbourhood search
2.6.3 Adaptive large neighbourhood search
2.6.4 Hyperheuristics
2.7 Conclusions
3 The personnel rostering model
3.1 The differences between the standard academic models and the
extended generic model
3.2 The problem model
3.2.1 Schedule period
3.2.2 Schedule definitions
3.2.3 Schedule constraints
3.2.4 Schedule
3.3 Conclusions
4 Evaluation of constraints
4.1 Definitions and variables
4.2 Hard constraints
4.2.1 Single assignment start per day per employee
4.2.2 Schedule locks
4.2.3 Honour skill types
4.2.4 Defined assignments only
4.2.5 Overlapping shift types
4.3 Soft constraints
4.3.1 Rest times between shift types
4.3.2 Employee skill type penalties
4.3.3 Coverage constraints
4.3.4 Collaboration
4.3.5 Counters
4.3.6 Series
4.3.7 Successive series
4.3.8 Absence request
4.4 Constraint evaluation over multiple schedule periods
4.5 Conclusions
5 Quantitative measures of problem properties
5.1 Minimum number of required assignments
5.2 Minimum number of required and free assignments
5.3 Tightness ratio
5.4 Conclusions
6 Nurse rostering benchmarks
6.1 The KAHO benchmarks
6.2 Conclusions
7 Solution methods
7.1 Initialisation Method
7.2 Hyperheuristic approach
7.3 Tabu Search
7.4 Variable neighbourhood search
7.5 Adaptive large neighbourhood search
7.6 Neighbourhoods and heuristics
7.6.1 Neighbourhoods
7.6.2 Heuristics
7.7 Conclusions
8 Experiments
8.1 Experimental settings
8.2 Experimental results
8.2.1 Results on VNS and ALNS variants
8.2.2 Results on hyperheuristics variants
8.3 Conclusions
9 Conclusions
9.1 Future research
ISBN: 978-94-6018-490-1
Publication status: published
KU Leuven publication type: TH
Appears in Collections:Computer Science Technology TC, Technology Campuses Ghent and Aalst
Technologiecluster Computerwetenschappen
Computer Science, Campus Kulak Kortrijk
Informatics Section

Files in This Item:
File Status SizeFormat
phdBurakBilgin.pdf Published 2761KbAdobe PDFView/Open


All items in Lirias are protected by copyright, with all rights reserved.