Dynamic Control of a Make-to-Order Manufacturing System under Throughput Time Constraints: An Effective Computational Method in the High-Dimensional Case

Abstract

We consider dynamic control of a make-to-order production system that serves price- and delay-sensitive customers. The production system is modeled as a multiclass queueing network where the customers' delay sensitivity is modeled via throughput-time constraints. To maximize her profit while meeting these throughput time constraints, the system manager uses three different types of control: i) dynamic pricing decisions; ii) dynamic order-of-service (scheduling) decisions; and iii) outsourcing decisions upon arrival of new orders. We approximate this problem by a multi-dimensional Brownian control problem in heavy traffic, and develop a neural network-based computational method tailored for high-dimensional problems. By interpreting the solution in the context of the original model, we propose an integrated dynamic pricing, outsourcing, and scheduling policy. The effectiveness of this policy is demonstrated through a simulation study of two test problems drawn from the literature.