Reducing domains for search in CLP(FD) and its application to job-shop scheduling

Abstract

This paper describes a generalization of the “labelling” search strategy and its application to scheduling problems. The assignment of a value to the selected variable is replaced by reduction of the domain of the variable. This strategy can be applied for solving problems modelled in CLP(FD).We discuss the application of this domain-reducing strategy to the well-known problem of 10×10 job-shop scheduling. The computation results obtained using this strategy show its advantages. A good solution, coming to within 3% of the optimal solution, is generated in less than 1 second, using only a simple heuristics for variable selection and domain reduction. Compared with the standard strategy, the domain-reducing strategy exhibits more robustness with respect to the given planning horizon. Assuming that there is a choice of machines, a good solution can be generated with the same strategy, deviating from the optimal solution by less than 4%.Our experience has shown that the domain-reducing strategy is suitable as a basic search strategy for solving job-shop problems by means of CLP, allowing good (near-to-optimum) solutions to be computed fast.