Simulation-based meta-optimization in vendor managed inventory systems
MetadataShow full item record
Vendor-managed inventory (VMI) is a supply chain coordination program in which the manufacturer takes full control of inventory management for retailers. The retailer does not have the responsibility of monitoring inventory or placing orders. Although the VMI philosophy has been embraced in a number of industries, there are few research works that attempt to develop analytical models for it because of the complexity in modeling. Simulation has been used to model VMI systems in the literature, but none of the existing works seek optimal inventory policies. Hence, a methodology based on simulation optimization has been developed here to obtain the optimal inventory policies in VMI systems. The solution methodology for the simulation model is based on two paradigms: a dynamic slave paradigm, which optimizes the inventory replenishment policy for the retailers, and a static master paradigm, which optimizes the inventory replenishment policy for the distribution center (DC). The slave problem is modeled as a semi-Markov decision problem. Approximate dynamic programming, also called reinforcement learning, is used to optimize the inventory replenishment policies for the retailers. For the master paradigm, a genetic algorithm is applied to optimize the ( Q, R ) policy for the inventory replenishment at the DC. Numerical examples are presented which show that the methodology based on reinforcement learning and genetic algorithms outperforms some well-known benchmarking algorithms.