A Column Generation Scheme for Faculty Timetabling

In this paper we deal with the problem of building a timetable for the courses of a university faculty. We propose an integer linear programming approach based on column generation. Each column is associated with a weekly timetable of a single course. The constraints referring to classroom occupancy and the non-overlapping in time of courses are in the integer linear programming matrix. The constraints and preferences related to a single course timetable are embedded in the column generation procedure. Generating a column for a course amounts to selecting the currently best time Slots in the week. The interaction between the column generation procedure and the branch-and-bound method is also discussed. Some computational results are shown.