The university course timetabling problem deals with the task of scheduling lectures of a set of university courses into a given number of rooms and time periods, taking into account various hard and soft constraints. The goal of the International Timetabling Competitions ITC2002 and ITC2007 was to establish models for comparison that cover the most frequently found use cases. Our model, motivated by a project with University College London (UCL), builds on the standard model from track 3 of ITC2007. Compared to the standard model from the literature, we cover several new constraints and extra features. For example, we expand the ITC2007 framework to generate a timetable for several weeks of the term instead of only one and introduce the corresponding timetable regularity metric, which measures the consistency of time and room assignments for a course throughout the term. We suggest an Integer Linear Programming approach for solving this expanded timetabling problem and introduce a corresponding new benchmark library. Finally we conduct computational experiments and discuss the results obtained with respect to solution quality and practical suitability for UCL.
Operations Research Proceedings 2016, accepted.