Implications of Agent-based Supply Chain Games.pdf

Implications of Agent-based Supply Chain Games Tracy Mullen1, Shuang Sun1, Moti Levi2, John Bagby1, Viswanath Avasarala1, John Yen1 School of Information Sciences and Technology1, SMEAL College of Business2 Pennsylvania State University University Park, PA 16802 Abstract – The increased prevalence of network-enabled supply chains and out-sourcing of business processes suggests a stronger role for simulation tools, such as multi-agent systems, in supply chain management. We report on a new supply chain management game in the 2003 Trading Agent Competition and the design and experiences of our agent, PSUTAC. We discuss how using a shared mental model approach can help SCM designers attack the role of information flow in an uncertain market environment. We conclude with a discussion about future implications to SCM of such trading agent simulations. I. INTRODUCTION The recent trends towards out-sourcing of many business processes, and modularization of supply-chains, result in a growing need for review and revision of traditional supply chain management (SCM) tools. Such tools generally model decisions as being made by one centralized decision maker, rather than as a decentralized negotiation and decision-making process. At the same time, analytical models are limited in their ability to model complex, multi-firm, multi-dimensional relationships. New simulation tools, including multi-agent systems, are starting to be investigated. Multi-agent system design meshes well with modeling supply chain networks, as it inherently assumes that agents have their own goals, which may be anywhere from pure self-interest to cooperative, thus allowing more freedom of analysis compared to traditional simulation or analytical tools. The 2003 Trading Agent Competition added a Supply Chain Management game (TAC/SCM) [1] to stimulate interest and research in this area. We discuss the design of our agent, PSUTAC, our experiences in this game, an