Combining Incomplete Search and Clause Generation: An Application to the Orienteering Problems with Time Windows
Résumé
In this paper, we present a hybrid optimization architecture which combines on one side incomplete search processes that are often used to quickly find good-quality solutions to large-size problems, and on the other side clause generation techniques that are known to be efficient to boost systematic search. In this architecture, clauses are generated once a locally optimal solution is found. We introduce a generic component to store these clauses generated step-by-step. This component is able to prune neighborhoods by answering queries formulated by the incomplete search process. We define three versions of this clause basis manager and then experiment with an Operations Research problem known as the Orienteering Problem with Time Windows (OPTW) to show the efficiency of the approach.