Date on Master's Thesis/Doctoral Dissertation


Document Type

Doctoral Dissertation

Degree Name

Ph. D.


Industrial Engineering

Committee Chair

Heragu, Sunderesh S.

Author's Keywords

Distribution; Agent-based simulation; Inventory; Supply chain; Production; Benders decomposition


Business logistics; Production management; Manufacturing processes; Industrial efficiency


In today's competitive business environment, companies face enormous pressure and must continuously search for ways to design new products, manufacture and distribute them in an efficient and effective fashion. After years of focusing on reduction in production and operation costs, companies are beginning to look into distribution activities as the last frontier for cost reduction. In addition, an increasing number of companies, large and small, are focusing their efforts on their core competencies which are critical to survive. This results in a widespread practice in industry that companies outsource one or more than one logistics functions to third party logistics providers. By using such logistics expertise, they can obtain a competitive advantage both in cost and time efficiency, because the third party logistics companies already have the equipment, system and experience and are ready to help to their best efforts. In this dissertation, we developed an integrated optimization model of production, inventory and distribution with the goal to coordinate important and interrelated decisions related to production schedules, inventory policy and truckload allocation. Because outsourcing logistics functions to third party logistics providers is becoming critical for a company to remain competitive in the market place; we also included an important decision of selecting carriers with finite truckload and drivers for both inbound and outbound shipments in the model. The integrated model is solved by modified Benders decomposition which solves the master problem by a genetic algorithm. Computational results on test problems of various sizes are provided to show the effectiveness of the proposed solution methodology. We also apply this proposed algorithm on a real distribution problem faced by a large national manufacturer and distributor. It shows that such a complex distribution network with 22 plants, 7 distribution centers, 8 customer zones, 9 products, 16 inbound and 16 outbound shipment carriers in a 12-month planning period can be redesigned within 33 hours. In recent years, multi-agent simulation has been a preferred approach to solve logistics and distribution problems, since these problems are autonomous, distributive, complex, heterogeneous and decentralized in nature and they require extensive intelligent decision making. Another important part in this dissertation involved a development of an agent-based simulation model to cooperate with the optimal solution given by the optimization model. More specifically, the solution given by the optimization model can be inputted as the initial condition of the agent-based simulation model. The agent-based simulation model can incorporate many other factors to be considered in the real world, but optimization cannot handle these as needed. The agent-based simulation model can also incorporate some dynamics we may encounter in the real operations, and it can react to these dynamics in real time. Various types of entities in the entire distribution system can be modeled as intelligent agents, such as suppliers, carriers and customers. In order to build the simulation model more realistic, a sealed bid multiunit auction with an introduction of three parameters a, ß and y is well designed. With the help of these three parameters, each agent makes a better decision in a simple and fast manner, which is the key to realizing real-time decision making. After building such a multi-agent system with agent-based simulation approach, it supports more flexible and comprehensive modeling capabilities which are difficult to realize in a general optimization model. The simulation model is tested and validated on an industrial-sized problem. Numerical results of the agent-based simulation model suggest that with appropriate setting of three parameters the model can precisely represent the preference and interest of different decision makers.