Modelling and solving the thesis defense timetabling problem∗

The thesis defense timetabling problem consists in composing the suitable committee for a set of defense sessions and assigning each graduation candidate to one of the sessions. In this work, we define the problem formulation that applies to some Italian universities and we provide three alternative solution methods, based on Integer Programming, Constraint Programming and Local Search, respectively. We also develop a principled instance generator, in order to expand the set of available instances. We perform an experimental analysis and we compare our solvers among themselves, using a testbed composed of both real-world and artificial instances. Even though there is no dominant method, the outcome is that Integer Programming gives the best average results, with Local Search being second, and Constraint Programming last on our testbed. All data is available on the web for verification and future comparisons.