Towards the Benchmarks for Scheduling Problems

The aim of this paper is not to offer any definite proposal for the scheduling competition, but to discuss some of the key issues to be considered in such an endeavour. The discussion will refer to three types of scheduling problems: production scheduling, employee scheduling (the discussion will be focused on nurse rostering being a very complex and highly constrained scheduling problem) and university timetabling. The reasons for choosing these types of scheduling problems are twofold: they are widely studied in the scheduling community and have led to a wealth of the scheduling models and algorithms, and the author has experience in solving these problems. However, the competition could include other types of scheduling problems, such as project scheduling, etc. First, a brief description of each of these types of problems will be given, together with available benchmark problems. After that, the following issues will be addressed: (a) is there a need for a common formal representation of different scheduling problems, (b) how to perform the evaluation of the scheduling algorithms, (c) is there a need for both randomly generated problem instances and real-world ones, and (d) what conclusions could we draw after the competition. The paper will also draw upon the experiences gained in the International Timetabling Competition organized by the Metaheuristic Network and sponsored by Patat (a series of conferences on Practice and Theory of Automated Timetabling) that was held in 2003. More information about this competition can be found on the Web page http://www.idsia.ch/Files/ttcomp2002/. The 2 International Timetabling Competition will be organised in the near future.