Multi-Neighbourhood Local Search for Course Timetabling

One of the most critical features for applying local search to timetabling problems is the definition of the neighbourhood structure. In fact, for most common timetabling problems, there is more than one neighbourhood structure that is sufficiently natural and intuitive to deserve investigation. The reason for considering combination of diverse neighbourhood is related to the diversification of search needed to escape from local minima. In fact a solution that is a local minimum for a given definition, is not necessarily a local minimum for another one, and thus an algorithm that uses both has more chances to move toward better solutions. There are actually many ways to combine different neighbourhoods. In this work, we investigate the following three.