Logistics with competing users

Abstract

This dissertation addresses a class of fundamental logistical problems where two or more potential users (or players) compete for a common set of resources. Each user has a criterion (cost or performance requirement) that he/she wishes to optimize. The users' criteria are often in conflict, that is, choosing a decision that optimizes one user's criterion may not also optimize the criteria of others. How should the resources be utilized to satisfy the user demands? In this dissertation, optimization and game theoretical models are employed to examine the equilibrium points and efficiently find the frontier of non-dominated solutions to three logistics problems with competing users: (1) single machine scheduling, (2) network resource allocation and (3) assignment to multiple servers (queues). New cooperative game theoretic methods are developed to negotiate on the Pareto frontier. In addition, a Stackelberg leader-follower game framework is introduced in a queueing system which includes both competitive users and competitive servers. The existence of a unique equilibrium is shown. The models and methodologies developed in the dissertation can be applied in many areas, such as Internet pricing, scheduling resources among competitors, network routing of users' requirements, analysis of competitive market, etc.