Hierarchical decision making with supply chain applications

Hierarchical decision making is a decision system, where multiple decision makers are involved and the process has a structure on the order of levels. It gains interest not only from a theoretical point of view but also from real practice. Its wide applications in supply chain management are the main focus of this dissertation.The first part of the work discusses an application of continuous bilevel programming in a remanufacturing system. Under intense competitive pressures to lower production costs, coupled with increasing environmental concerns, used products can often be collected via customer returns to retailers in supply chains and remanufactured by producers, in orderto bring them back into “as-new” condition for resale. In this part, hierarchical models are developed to determine optimal decisions involving inventory replenishment, retail pricingand collection price for returns. Based on the simplified assumption of a single manufacturer and a single retailer dealing with a single recoverable item under deterministic conditions,all of these decisions are examined in an integrated manner. Models depicting decentralized, as well as centralized policies are explored. Analytical results are derived and detailed sensitivity analysis is performed via an extensive set of numerical computations.In the second part of this dissertation, a discrete bilevel problem is illustrated by investigating a biofuel production problem. The issues of governmental incentives, industry decisions of price, and farm management of land are incorporated. While fixed costs are natural components of decision making in operations management, such discrete phenomena have not received sufficient research attention in the current literature on bilevel programming, due to a variety of theoretical and algorithmic difficulties. When such costs are taken into account, it is not easy to derive optimality conditions and explore convergence properties due to discontinuities and the combinatorial nature of this problem, which is NP-hard. In order to solve this problem, a derivative-free search technique is used to arrive at a solution to this bilevel problem. A new heuristic methodology is developed, which integrates sensitivity analysis and warm-starts to improve the efficiency of the algorithm.