Improvements of constraint programming and hybrid methods for scheduling of tests on vehicle prototypes

• Published in: Constraints : an international journal Vol. 17; no. 2; pp. 172 - 203
• Main Authors: Limtanyakul, Kamol, Schwiegelshohn, Uwe
• Format: Journal Article
• Language: English
• Published: Boston Springer US 01.04.2012; Springer
• Subjects: Application; Applied sciences; Artificial Intelligence; Computer Science; Exact sciences and technology; Mathematical programming; Operational research and scientific management; Operational research. Management science; Operations Research/Decision Theory; Optimization; Mixed integer linear programming; Benders decomposition; Automotive industry; Constraint programming; Mass production; Partition method; Linear programming; Mixed integer programming; Scheduling; Temporal constraint; Modeling; Search algorithm; Constrained optimization; Genetic algorithm; Automobile industry; Heuristic method; Memory capacity; Robustness; Planning
• ISSN: 1383-7133; 1572-9354
• DOI: 10.1007/s10601-012-9118-y

In the automotive industry, a manufacturer must perform several hundreds of tests on prototypes of a vehicle before starting its mass production. Tests must be allocated to suitable prototypes and ordered to satisfy temporal constraints and various kinds of test dependencies. The manufacturer aims to minimize the number of prototypes required. We present improvements of constraint programming (CP) and hybrid approaches to effectively solve random instances from an existing benchmark. CP mostly achieves better solutions than the previous heuristic technique and genetic algorithm. We also provide customized search schemes to enhance the performance of general search algorithms. The hybrid approach applies mixed integer linear programming (MILP) to solve the planning part and CP to find the complete schedule. We consider several logical principles such that the MILP model can accurately estimate the prototype demand, while its size particularly for large instances does not exceed memory capacity. Moreover, the robustness is alleviated when we allow CP to partially change the allocation obtained from the MILP model. The hybrid method can contribute to optimal solutions in some instances.