The supply chain is a network of suppliers, factories, warehouses, distribution centers and retailers through which raw materials are acquired, transformed and delivered to the customer. Supply chain management is the decision making that optimizes supply chain performance. In today's competitive business environment industry is recognizing the importance of efficient supply chain management.
HP management has recognized that its performance filling orders will cause it to win or lose the competitive battle. (Lee and Billington)
It has become increasingly apparent that limits in achieving this lie not just in labour or capital, but the availability/accessibility of information and the ability to effectively coordinate both decisions and actions.
In a distributed domain such as the supply chain, where any local decision may have widespread effects, a key to meeting the future challenges is the development of next generation information and management systems. These must support coordinated problem solving and decision making in an integrated supply chain. A promising approach is the use of multi-agent systems.
An agent can be seen as a piece of software that is significantly autonomous, goal-oriented and entrusted in performing its functions and that operates globally on networks by relying on application-independent high-level communication and interaction protocols with other "agents". The COOrdination Language (COOL) is developed at the Enterprise Integration Laboratory (EIL) of the University of Toronto. COOL provides constructs for defining agents and protocols for coordination among the agents, so-called conversations. The conversations allow the agents to share information and coordinate their problem solving through message passing based on KQML  (the Knowledge Query and Manipulation Language).
This thesis will investigate the use of agent technology to improve performance within important aspects of supply chain management, with focus on customer service and inventory management. A supply chain model will be designed and implemented in COOL. Through simulation on different configurations of the model, the impact of information sharing and coordination on supply chain performance will be analyzed.
The thesis is related to the Integrated Supply Chain Management (ISCM) Project at the Enterprise Integration Laboratory at the University of Toronto.
The EIL research enables businesses to develop, manufacture, sell, deliver and support products and services with unprecedented speed, flexibility, quality and economy. This is achieved through the application of business practices and technologies that create a business infrastructure enabling the dissemination of information, coordination of decisions, and management of actions to and among people and systems within the organization and outside of it. EIL research explores the creation of Enterprise Integration concepts in a bi-directional manner, in that it is simultaneously theory and application driven; an underlying philosophy to this research is that solving real problems leads to breakthrough research. The theories that are being explored include: coordination theory, common sense enterprise modeling and agent-based enterprise information architectures. Applications include enterprise design, concurrent engineering and integrated supply chain management. (from Fox )
The objectives of the ISCM Project are:
As a part of the ISCM Project, a supply chain demonstrator was implemented. The supply chain demonstrator is a very simple model, which aims to show an example of a supply chain application implemented in COOL. The supply chain demonstrator was the starting point of this thesis. Rather than extending the demonstrator, it was decided to implement another, more realistic, supply chain model.
The objective of the thesis is to: Analyze, describe and design information and work-flows for supply chain coordination. The objective is divided into two main issues.
The first main issue is to study and evaluate the information flow in a supply chain demonstrator. This is interpreted to mean to study and evaluate the need for, and use of, information sharing in a supply chain model.
The second main issue is to extend supply chain agent functionalities and/or typologies to provide for improved information and work flow requirements. The term improved as used here, is interpreted to mean more consistent with a real life supply chain. As a whole, the second main issue of the thesis objective is interpreted to be: Extend the functionalities and/or typologies of the agents which are used in the current supply chain demonstrator to provide for information and work flow requirements that are more consistent with those of a real life supply chain.
Through the work with the thesis it has become evident the second main issue will precede the first. First the PMC Model is designed and implemented, next the effects of different strategies for information sharing and coordination are studied and evaluated in this context.
Figure: The Chapters of the Thesis.
As can be seen from Figure , the thesis is divided into two chapters describing the practical work (in dark gray) and three chapters describing its theoretical background (in bright gray). The majority of the work has been put down in the design and implementation the Perfect Mini Computer Model, described in Chapter . Both modeling and implementation proved to be very time consuming. (That COOL (and also LISP) was unknown to the author prior to the start of the work contributed to this.)
The theoretical background starts with a presentation of supply chain management in Chapter . The chapter presents some important issues within supply chain management, focusing on customer service and inventory management, and how these are related to flexibility. The chapter also presents opportunities for future improvements, now focusing on integration and coordination. Finally the notions of modeling and simulation are presented.
Chapter deals with the second factor the thesis, the intelligent agent. The reader is familiarized with the notion of agency and agent theory, architecture, languages, and applications. The last section is directly based on the previous chapter, and gives a motivation for the use of agent technology to construct supply chain information and management systems.
The COOrdination Language, which is the agent language used in the practical work, is described in Chapter . The chapter describes the features of the language: agents, conversation classes, conversation rules, continuation rules, and conversation managers, as well as the graphical user interface it provides.
The two sides of theoretical background, supply chain management and agents, lead up to the first of the two chapters describing the practical work. Chapter is dedicated to a first of two models implemented using COOL, the Simple Model. The Simple Model was implemented for the author to be familiarized with the tools provided by COOL, as well as the behavior of a simple supply chain model. Many of the solutions used for the Simple Model are in principle reused for the PMC Model. It may therefore be useful for the reader to look through the chapter, though higher priority should be given to the next chapter.
Chapter presents the PMC Model, first independently of COOL, then as it was implemented. Further, some results of the simulations that were run on the model are described and analyzed. The conclusion section of the chapter will, among others, compare the experience and results obtained, with the statements made in the motivation section of chapter .
It has not been the aim, when designing the model, to find optimal solutions locally, but rather to provide a test-bed for different coordination strategies. The focus has, in other words, been on designing for coordination and information sharing among the entities of the supply chain, rather than to give each entity powerful problem solving tools.
Due to the time constraints, it was important to restrict the focus when designing the PMC Model. The focus has been put on customer satisfaction and inventory values, particularly the raw product inventories (RPI). This is also reflected in the theoretical background in Chapter .