MASCF: A generic process-centered methodological framework for analysis and design of multi-agent supply chain systems

Abstract

Multi-agent systems (MAS) are becoming popular for modeling complex systems such as supply chains. However, development of multi-agent systems remain quite involved and extremely time consuming. Currently, there exist no generic methodologies for modeling supply chains using multi-agent systems. In this research, we propose a generic process-centered methodological framework, Multi-Agent Supply Chain Framework (MASCF), to simplify MAS development for supply chain (SC) applications. MASCF introduces the notion of process-centered organization metaphor, and creatively adopts Supply Chain Operations Reference (SCOR) model to a well-structured generic MAS analysis and design methodology, Gaia, for multi-agent supply chain system (MASCS) development. The popular Tamagotchi case was designed and analyzed using MASCF. The validity of the framework was established by implementing MASCF output of Tamagotchi SC using the Java Agent DEvelopment Framework (JADE).

Introduction

Supply chain management (SCM) has emerged as an invaluable strategic initiative providing competitive advantage for enterprises in the market place. Gartner-Dataquest Report (2005) defines SCM as a business strategy of integrating business processes from a supplier’s supplier to a customer’s customer that creates and fulfills the market’s demand for goods and services, and optimizes the flow of products, services and information. Chopra and Meindl (2006) define supply chain to be the dynamic involving constant flow of information, products, and funds between different stages that perform different processes while interacting with the other stages. Christopher (1998) defines supply chain as a network of organizations linked through upstream and downstream processes that add value to the ultimate customer through products and services. It is clear from these definitions that the sheer scope of a supply chain makes its efficient management a complex task. In today’s global marketplace, as more and more firms embrace SCM, individual firms no longer compete as independent enterprises but rather as integral parts of supply chains (Lambert et al., 1998, Min and Zhou, 2002). The constant battle for supremacy is no longer between an enterprise and its competitors, but between the supply chain of the enterprise and those of its competitors (Baatz, 1995, Taylor, 2003). The success of any enterprise, accordingly, depends on its ability to integrate and coordinate the intricate network of business processes among its supply chain partners efficiently (Drucker, 1998, Lambert and Cooper, 2000). It follows that the ultimate success of an enterprise is not derived independent of, but coupled with the destiny of its supply chain.

Traditional supply chain modeling and management involves the application of: optimization (e.g., Arntzen et al., 1995, Beamon, 1998, Goetschalckx et al., 2002, Lee and Billington, 1995), mathematical models (e.g., Anupindi and Bassok, 1999, Cachon, 2003, Cachon and Fisher, 2000, Lee and Whang, 1999), simulation (e.g., Bhaskaran, 1998, Chan et al., 2002, Holweg and Bicheno, 2002, Petrovic, 2001, Terzi and Cavalieri, 2004), system dynamics (e.g., Angerhofer and Angelides, 2000, Higuchi and Troutt, 2004, Sterman, 1989), and others. These approaches usually employ a centralized decision-making treatment, and typically involve a single comprehensive model, under the assumption of information symmetry (every bit of information is known to every one else or at least available to the model builder/decision maker). Another trend in this area that is gaining prominence is the usage of a combination of tools in decision modeling, for example: simulation–optimization (Padmos, Hubbard, Duczmal, & Saidi, 1999). Latest developments in the computing and object-oriented technologies facilitate modularization and development of reusable objects leading to rapid development of models (e.g., Bagchi et al., 1998, Biswas and Narahari, 2004). In addition, Graphical-User-Interface (GUI) based technologies available today simplify model description through drag-and-drop features. All these advancements play a crucial role in developing and solving bigger and more realistic models quickly. Traditional modeling techniques are quite suitable for modeling supply chain decisions within a single enterprise. Organizations have been applying these techniques for several decades leading to higher efficiencies. Given that intra-enterprise modeling helped improve the efficiencies a great deal, modeling of inter-enterprise issues for SC integration are crucial for further large-scale improvements. Considering the fact that most of the supply chains involve enterprises with independent ownerships (requiring the ability to model information asymmetry and distributed/decentralized mode of controls), applicability of the traditional modeling approaches is quite limited and indeed unrealistic. The latest developments in the modeling technology, agent-based systems, and multi-agent systems for example, are quite promising for such modeling situations. They are best suited to handle issues of information asymmetry, decentralized and distributed decision-making, and modeling inter-enterprise issues.

Agent technologies are an offshoot of “Distributed Artificial Intelligence (DAI)” that has a long-term goal of developing mechanisms and methods that enable agents to interact as well as human beings, or even better (Weiss, 1999). Autonomous agents and multi-agent systems represent a new way of analyzing, designing, and implementing complex software systems (Jennings, Sycara, & Wooldridge, 1998). They are expected to pioneer a revolutionary paradigm shift in software systems modeling and engineering (Zambonelli & van Dyke Parunak, 2003). Multi-agent systems can be used to model any phenomenon, scientific or behavioral, in order to study the underlying dynamics of complex systems such as supply chains very effectively. Agents can be modeled to represent organizations, functions, resources, and even human beings. They have the ability to incorporate within, some of the existing modeling approaches (e.g., optimization, simulation, game theory), making them more powerful. They can also be made to learn with the help of artificial intelligence tools and techniques, leading to “intelligent” agents. This, however, does not mean that “agents” are the panacea for all the modeling issues. Wooldridge and Jennings, 1998, Wooldridge and Jennings, 1999 while emphasizing that intelligent agent and multi-agent systems can potentially play a significant role in complex and distributed systems engineering, warn of avoiding potential pitfalls in engineering industrial strength agent-oriented software systems. Multi-agent supply chain literature is scant in terms of richness of real-world applications and implementations. While this is expected due to being a relatively new and upcoming field, there exist several other reasons too. For example, the modeling standards are still being evolved and the infrastructural support in terms of tools and methodologies are still in their nascent stage of development. Given that the development of multi-agent systems is quite involved and time consuming, the necessary infrastructure support plays an extremely important role in their large-scale adoption. We argue that such a support simplifies the model building process leading to the proliferation of real-world implementations.

Our research accordingly focuses on simplifying multi-agent system development, in particular, for supply chain applications. We propose Multi-Agent Supply Chain Framework (MASCF), a generic process-centered methodological framework, towards this goal. MASCF is designed to facilitate and simplify the analysis and design phases of the development. Instead of developing yet another methodology afresh from ground zero, we design our framework around already well established models/methodologies that become its key elements. The framework introduces the notion of process-centered organization metaphor, and creatively adopts a generic process-standard for supply chain description (Supply Chain Operations Reference model, SCOR) to a well-structured generic methodology for multi-agent system development (Gaia). Since SCOR and Gaia are the key elements, MASCF is a generic tool widely applicable, and practical for modeling supply chains through multi-agent systems.

This paper is structured as follows. Pertinent literature is reviewed in Section 2. Section 3 describes the key elements of the framework. MASCF is introduced in Section 4 along with its scope and limitations. The operating mechanics of the framework are discussed in detail in Section 5. The framework is validated with the help of a case study, Tamagotchi, and its details are provided in Section 6. Finally, Section 7 summarizes research contributions and identifies some of the extensions.