Multi-neighbourhood Local Search with Application to Course Timetabling

A recent trend in local search concerns the exploitation of several different neighbourhood functions so as to increase the ability of the algorithm to navigate the search space. In this work we investigate the use of local search techniques based on various combinations of neighbourhood functions, and we apply it to a timetabling problem. In particular, we propose a set of generic operators that automatically compose neighbourhood functions, giving rise to more complex ones. In the exploration of large neighbourhoods, we rely on constraint techniques to prune the list of candidates. This way, we are able to select the most effective search technique through a systematic analysis of all possible combinations built upon a set of basic, humandefined, neighbourhood functions. The proposed ideas are applied to a practical problem, namely the socalledCourse Timetabling problem. Our algorithms are systematically tested and compared on real-world instances. The experimental analysis shows that neighbourhood composition leads to much better results than traditional local search techniques.