Job scheduling integrated with material ordering: decision-dependent stochastic programming and information relaxation dual bounds

Abstract

We investigate an integrated job scheduling and material ordering problem arising from one-of-a-kind production, accounting for endogenous uncertainties in material consumption. The integrated problem is formulated as a multistage stochastic program incorporating a set of linear non-anticipativity constraints, while model reduction properties are also carefully identified. To address the computational complexity, we introduce a rolling horizon procedure to generate quality feasible policies in a sequential and dynamic manner. Our primary contribution lies in the development of a penalized information relaxation duality approach, which constructs a dual bound for the proposed multistage stochastic program. This innovative method is convenient for Monte Carlo simulation and parallel resolution, offering promising potential for future research. Computational results validate the efficiency and effectiveness of our approaches in minimizing total costs compared with existing methods. Additionally, sensitivity analyses are conducted to identify factors influencing the computational performance of our algorithms, including the planning horizon, uncertainty variance, sample size, and network structure.