Files
Abstract
Problems originated from multi-tier networks are central to the field of OM/OR. Over the years, multi-tier networks has only gotten more complex. On one hand, companies today are building more local distribution centers and opening up more retail shops to serve regional demand. On the other hand, companies are competing in goods delivery time to take customer service to the next level. In this dissertation, we study inventory and fulfillment policies in a two-tier network, where the upper tier consists of one central warehouse or regional distribution center (RDC), and the bottom tier consists of multiple retailers or front distribution centers (FDCs).
Classic literature on multi-echelon inventory policy assumes a steady state that a manufacturer is always able to place an order and receive it within a reasonable time frame. In Chapter 2, we first assume the same steady state, and we consider the problem of minimizing the long-run cost of a two-tier network with multiple retailers and with expediting. The features of multi-location and expediting are omnipresent and critical to supply chain networks in practice. Furthermore, due to the recent pandemic, we later drop the steady state assumption in Chapter 2 and study the problem of allocating limited inventory across the two-tier network, when the manufacturer is unable to receive external supplies.
Chapter 3, on the other hand, is solely concerned with fulfilling orders in a two-tier network. We assume no inventory replenishment can happen during a fulfillment period, we allow orders to consist of multiple items, and we allow orders to be split into multiple packages for fulfillment. Because an order may contain more than one item, the decision maker needs to efficiently decide what distribution centers to use to fulfill what part of an order. Chapter 3 studies a widely-implemented myopic policy in such a setting and evaluates the policy's performance in competitive analysis. Chapter 3 also studies a linear program rounding policy and a delay policy, theoretically and numerically.