Tabu Search for a Job Scheduling Problem
Résumé
We consider a generalization of the graph coloring problem, where variable costs are encountered for assigning a color to a vertex, and incompatibility costs for giving two vertices the same color. We state this problem in terms of single-period jobs (vertices), each of which has to be assigned a time period (color). We propose a tabu search heuristic, as well as an adaptive memory algorithm, and compare them with other heuristics on large instances, and with an exact method (integer linear program) on small instances. We also show how our model permits to incorporate precedence constraints between jobs.