《2016 A Compact Optimization Approach for Job Shop Problems》.pdf

Regular Papers A Compact Optimization Approach for Job-Shop Problems Giuseppe Lancia, Franca Rinaldi, Paolo Serafini University of Udine, Dept. of Mathematics and Computer Science, Via delle Scienze 206, Udine, Italy, {lancia, rinaldi, serafini}@dimi.uniud.it Abstract: In this paper we propose a time-indexed IP formulation for job-shop scheduling prob- lems. We first introduce a model with variables associated to job scheduling patterns and constraints associated to machine capacities and to job assignments. The exponential number of variables calls for a column generation scheme which is carried out by a dynamic programming procedure. How- ever, the column generation process is time-consuming. Therefore we derive an equivalent compact formulation of the model, whose dual is a flow problem with side constraints. This problem can be solved faster and fits easily into a branch-and-bound scheme. Keywords: Machine Scheduling, Shop-Floor Scheduling, Compact optimization. 1 Introduction The job-shop problem has been studied extensively in the last three decades. Although there are good heuristic procedures based on local search (for instance the celebrated shifting bottleneck procedure [1] is a local search with a very large neighborhood), the performance of exact procedures has never been satisfying. The hardness of the job-shop problem, even with respect to other NP- hard problems, has been soon recognized [6]. In this paper we introduce a time-indexed formulation for the general job-shop problem, based on an ILP model with column generation. In our model part of the combinatorial structure is taken care of by the constraints and part is embedded in the matrix columns. By this feature, typical of ILP formulations with column generation, the LP relaxation provides a strong lower bou