Combining exact and (meta)heuristic methods for problems of combinatorial optimisation.

We aim to perform research in the field of Hyper-Heuristics where one develops problemindepdent approximative mechanisms to tackle a certain problem. Currently the research field is quite sparse.

A recent survey (Burke et al.) describes todays problems with hyper-heuristics: a representative test bench is missing together with formalism and a way to gives guarantees on the quality of calculated solutions.

In this project we will tackle most of these issues with the following milestones:
1. The implementation of a large, representative and diverse testbench
2. The automatic generation of low-level heuristics based on a problem description
3. The realisation of a hyper-heuristics who takesas input the generated heuristics together with a concrete problem instance. We want to perform research on the possible guarantees a hyper-heuristic can give.
4. Injection of problem specific knowledge in order to speed up convergence towards the real solution or give better guarantees.
5. The implementation of a parallel hyper-heuristic.
6. Automatic problem specialisation of a hyper-heuristic together with a set of low-level heuristics