MACE-SCM : استدلال مبتنی بر مورد و چند عاملی مکانیسم همکاری برای مدیریت زنجیره تامین تحت عدم قطعیت عرضه و تقاضا

The importance of collaboration in the supply chain has led scholars to suggest diverse approaches for problems in the collaboration process. Questions still remain about which method is best when coordinating and sharing information in the presence of various supply and demand uncertainties. Hence, this paper aims to propose an integrated framework based on multi-agent collaboration and case-based reasoning that can resolve various collaboration issues in the supply chain. To show the framework’s feasibility, we implemented a prototype system: MACE-SCM. MACE-SCM provides more flexible and extensible solutions to help address emerging uncertainties. The validation results reveal the importance of intensive collaboration for maximum efficiency in the supply chain.

Firms today increasingly consider supply chain management (SCM) to be a major vehicle to gain a competitive advantage in turbulent markets. While firms have traditionally acted as sole economic entities in the market, they have begun to form strategic alliances with other firms, integrating their business processes, and consolidating their resources. According to the Global Supply Chain Forum, SCM is defined as … the integration of key business processes from end user through original suppliers that provides products, services, and information that add value for customers and other stakeholders (Lambert & Cooper, 2000). The advancement of information technology (IT) has allowed firms that participate in SCM to share information across organizational boundaries, bringing about substantial performance increases. For example, the collection of sales information at the point-of-sale and the sharing of that information via an electronic data interchange (EDI) have lowered costs in the ordering processes. Supply chain scholars have championed various complementary perspectives in order to resolve problems in collaboration and information sharing, including optimization-, simulation-, and multi-agent-based. Prior research focused primarily on optimization-based techniques and mathematical modeling of operational aspects of information sharing (Maturana & Norrie, 1997). Management Science/Operations Research (MS/OR) researchers have used this approach extensively to identify optimal solutions for given situations subject to specific assumptions. This approach is strong in addressing focused sets of problems, such as inventory management, logistics optimization, and production scheduling. Simulation-based approaches allow dynamic modeling of behaviors of supply chain firms with varying degrees of constraints and policies, dealing with diverse contingencies caused by supply and demand uncertainties. However, they cannot generate the design itself, and can only run models with pre-specified parameters and conditions (Harrison, 2001). Scholars have recently begun to focus on multi-agent-based approaches to address collaboration and information sharing problems (Swaminathan, Smith, & Sadeh, 1998). Supply chain firms are typically modeled as software agents, which pursue their own goals under certain constraints. Prior research has addressed diverse aspects in the supply chain, but questions still remain concerning the best method of addressing and resolving collaboration and information sharing problems. Traditionally, scholars have focused on the problem domain in which supply and demand uncertainties are low. In this context, the best strategy is to implement “efficient supply chains” (Lee, 2000 and Lee et al., 2000) by lowering costs. However, a supply chain may experience uncertainty both in a high supply and in a high demand. In high supply uncertainty, the supply chain suffers an evolving supply process in which manufacturing technology is emerging, and the supply base is unstable. The traditional approaches may have limited applicability in this context, because calculating analytical solutions are prohibitive—even impossible—as these uncertainties create increased complexities, resulting in a model that becomes overly complicated (Shapiro, 2001). The present research asks the following question: How can multi-agent and case-based reasoning be applied to facilitate collaboration and information sharing in the presence of high supply and demand uncertainties? We propose a framework based on a combination of multi-agent and case-based reasoning (MACE-SCM). This unique combination provides a model with several advantages. The multi-agent structure allows us to easily model different components in the supply chain. It also provides a flexible structure in handling emerging situations. For example, it can easily reflect on any changes in the agents’ roles or the structure of the supply chain. In addition, CBR mechanisms can be used to model different levels of collaboration; for example, intensive collaboration in which the partners share sensitive marketing information by tracking the history of problems and solutions. We first design three levels of collaboration among the partners: Autonomy, Integration, and Enhanced Integration. At the Autonomy level, the agents do not collaborate at all. At the Enhanced Integration level, the agents collaborate extensively; also, a central coordinator agent, MACE, which is equipped with the case base, provides additional support. We then expose the three to various uncertainties. The uncertainty contingencies considered are: (1) demand uncertainty; (2) supply uncertainties with regard to lead time, production capacity, change in the suppliers, and risk pooling. The validation results using six experiments demonstrate the importance of intensive collaboration for maximum performance. This paper is organized as follows: (1) briefly review existing literature on supply chain collaboration and information sharing; (2) present the architecture and detailed mechanisms of the model; (3) describe the implementation details and validation results of the prototype system; and (4) conclude with a brief discussion of the model’s implications.

The objective of the present research was to develop a framework based on multi-agent and case-based reasoning to facilitate collaboration and information sharing in the presence of high supply and demand uncertainties. The suggested system, MACE-SCM, was successful in addressing different uncertainty situations, and thus, facilitating collaboration between the firms within the supply chain. The validation results suggest that the combination of CBR and multi-agent based coordination mechanisms can produce the optimal results for the supply chain, as compared to CBR alone. The findings highlight the importance of utilizing additional strategic information in the case base in addressing supply chain uncertainties. The contribution of the research can be summarized as follows. First, we provide a solution for collaboration within the supply chain under various uncertainties. Contrary to the existing MS/OR approach, our approach does not require extensive knowledge in formulating and running the model. The multi-agent structure is intuitive to formulate, and the case base can be easily built by monitoring past transaction records of a supply chain. The three different levels of decision support can be used to simulate the benefits of different types of relationships before the firm enters into a strategic alliance with its partner. Similarly, the six experimental situations assist managers in evaluating the effects of different uncertainties. Second, our framework has a flexible structure in handling emerging situations. The multi-agent framework can handle the structural changes of the model. The addition and deletion of a firm can be addressed by modifying the sizes of agents. MACE can incorporate a new coordination mechanism, as well as modify the existing three mechanisms. In addition, the CBR structure can be used to handle contextual changes of the model. For example, given the existing structure of a supply chain, the emergence of a new uncertainty can be modeled by reconfiguring the case base to reflect new knowledge and emerging action strategies and policies. Last, our framework proves the usefulness of the central coordinator, MACE. As explained earlier, MACE was effectively used as a coordinator in adjusting global and local goals in supply chain. MACE avoids the behavioral problems that frequently take place when coordination conflicts occur among the agents. A few limitations of the research must be pointed out. First, a relationship type is operationalized by moving from strategic to transactional, and vice versa. In reality, the relationship may not be classified into an “on” or an “off” type, but has gray elements between the two opposite types. Thus, the current operation of relationship changes needs to incorporate additional factors and be fine-tuned to reflect the reality. Second, our experimental framework is grounded in the comparisons among three decision support levels, which do not include a base case. Our framework is more focused on showing the benefits of having multi-agent and CBR simultaneously. Thus, this concern can be best addressed by expanding the current framework to compare the experimental results with a MS/OR model based on linear/integer programming. We are expanding our research in that direction.