Finding near optimal solutions to the Thesis Defense Timetabling problem by exploiting symmetries

Abstract

Thesis Defense Timetabling is the problem of arranging meetings of graduation candidates and faculty members for evaluating the theses of the first ones. Several rules apply that guarantee that the evaluation process will be assigned to appropriate opponents based on their expertise, while work will be balanced among all faculty members. The problem is combinatorial in nature, and optimal solutions are hard to find and even harder to verify that they are indeed optimal. In this work we present an exact model of the problem that is solved using Constraint Programming. We augment this model by adding constraints that capture three types of symmetries that we identified. These additions make the models capable of finding near optimal solutions, in reasonable time, and this is demonstrated by running experiments on public datasets of Thesis Defense Timetabling problem instances. Several new best solutions and lower bounds are obtained. Our approach shows that exploiting intricate details of a problem can drive contemporary constraint programming solvers to optimal or near optimal solutions faster.