Solving scheduling problems from high-level models

Abstract

This is the summary of the author's PhD thesis supervised by Yves Deville and Pascal Van Hentenryck, and defended on 25 May 2010 at the Université catholique de Louvain in Louvain-la-Neuve. The thesis is written in English and is available from the author, or downloadable from http://becool.info.ucl.ac.be/files2/ thesis-monette.pdf. This work is about the gap between high-level modelling of scheduling problems and its efficient resolution. We propose to automatically classify models of scheduling problems, and to apply an appropriate search algorithm. Thanks to the strong separation between model and search, we propose also a simple way to create hybrid algorithms. A second part of the thesis deals with Constraint Programming approaches for two scheduling problems: The one-machine non-preemptive problem and the Just-In-Time Job-Shop problem. All parts are accompanied by experimental results showing their practicality.