Online Inventory Replenishment and Fleet Routing Decisions under Real-Time Information

Abstract

Logistics managers rely on increasingly sophisticated technologies to track demand and associated inventories, allowing rapid response to meet anticipated demand, avoid shortfalls while minimizing transportation and inventory carrying costs. This ability to respond gives rise to complex decision problems, characterized by combinatorial underlying problems under progressively unfolding demand. Real-time information also increases the ability to coordinate effectively inventory management and transportation service.

The advantages of coordinating inventory replenishment with vehicle routing decisions have long been recognized, giving rise to the inventory routing problem, which arises in the context of vendor-managed inventories. These typify an emerging class of collaborative logistics arrangements facilitated by information and communication technologies. The ability to coordinate inventory with routing decisions in real-time adds an important dimension to the problem. While fleet management decisions under real-time information have been studied extensively, coupling these with inventory replenishment decisions in real-time remains in the early stages of conceptualization and development. The main objective of this dissertation is to examine effectiveness of policies for managing inventories taking into consideration the interaction between inventory replenishment, retailer sequencing and transportation cost.

A major motivation for the online inventory routing problem is the presence of uncertainty about future consumption rates at different facilities. The possibility of updating plans on a continuous basis, based on real-time information about demand realizations makes possible decisions to modify the set and/or the sequence of subsequent facilities to be visited, diverting a truck from its current destination to visit a different facility, and adjusting amounts to be delivered to subsequent customers in the route.

This dissertation proposes two decomposition approaches, in which a simplified version of either the inventory-control or the routing side is solved first, and then that solution is used as a soft constraint when solving the other side. For each approach, different operational polices are proposed, reflecting different degrees of sophistication in terms of technology and optimization capabilities. These operational policies are based on a rolling-horizon framework, wherein new plans are repeatedly generated, based on updated information. Finally, the performance of proposed strategies is simulated and the impacts of using sophisticated real-time strategies are discussed.