Distribution requirements and compactness constraints in school timetabling

This paper provides the following contributions: First, distribution requirements for lessons of different lengths are modelled in a novel way by the use of the so-called multiple mode concept with mode identity constraints. Second, we show that several types of constraints may be modelled using the unifying framework of partially renewable resources. Among these constraints are: No class, subject, room, and teacher overlaps; class, subject, room, and teacher unavailabilities; compactness constraints; preassignment constraints; lectures to be given simultaneously; lunch breaks, etc. Third, we present two-phase parallel greedy randomized and genetic methods. Fourth, we provide an instance generator for the generation of a representative set of instances. Fifth, the generator along with a statistical model is used for a thorough experimental evaluation of the methods. Computational results show that the methods solve the instances investigated close to optimality.