- 學位論文
多重代理人技術與系統動力學方法整合應用於電子商業模擬系統之設計與實作
Integrating System Dynamic Modeling Technology and Multi-Agent Methodology for Applications to the Design and Implementation of e-Business Simulation Systems
摘要
本研究整合巨觀的系統動力學與微觀的多重代理人系統兩項塑模及模擬技術,應用於以經典的啤酒遊戲為基礎的ㄧ般供應鏈體系, 使作業層級、管理決策層級與策略層級的流程得以分離, 並嵌入現代供應鏈產業架構之SCOR標準, 以符合未來電子企業(e-business)網路資訊分享的特性需求。 同時我們以塑模及模擬系統建置的軟體工程角度, 以代理人軟體工程之Gaia方法論與UML圖呈現方式,遵循SCOR標準來擴充劃分細部流程比照原始啤酒遊戲程式碼,以比較可理解性、架構再用與擴充性等面向,而確認本方法及步驟適合於未來電子商務時代全球化運籌的商業複雜系統塑模與模擬; 在此成果基礎之上,進一步分別產生第三方物流及供應鏈網絡兩案例之塑模與模擬系統之分析設計與實作,來證明其可延伸擴充的能力以及發展性. 最後由塑模應用的角度, 將之應用於供應鏈管理教學之參與式模擬系統,可依據供應鏈模式個案的不同,讓多重代理人模型成為可選擇的劇本,並藉由智慧型代理人的支援來補足當參與者不足時,可指派智能或認知限制不同之代理人扮演個中角色,進行完整的個案模擬流程,讓人員體會完整決策與複雜互動的模擬過程.
並列摘要
This research combines two modeling and simulation technologies: System Dynamics of macro-level approach and Multi-agent system of micro-approach, and then applies to the general supply chain management based on the classic Beer Distribution Game. Our approach makes possible the separation of corporate management process into three levels: operational level、managerial level and strategic level, and extends with performance metrics in the modern supply chain industrial framework - SCOR. This method is expected to meet requirements of e-business characterized by network information sharing. Furthermore, from the software engineering viewpoint of system construction in modeling and simulating, we use the Gaia methodology from agent-oriented software engineering, with UML diagrams, and follow the SCOR standard to refine the processes. We then compare codes of the classic Beer Distribution Game program in terms of comprehensibility、reusability、expansibility, to confirm that the approach is suitable for modeling and simulating complex business systems operated with global logistics in the e-business era. With the foundation, we then produce system analysis and implementation of modeling and simulation cases in third-party logistics (3PLs) and supply chain networks, to further test its extensibility and ease of application development. As to applications of system modeling and simulation, we focus on the participatory simulation approach in teaching supply chain management. Based on the differences in targeted supply chain cases, multi-agent models are further equipped with a couple of optional case scenarios. When not enough human players are present to participate and initiate the simulation game, we can allocate versions of intelligent agents to play some designated roles in the scripted scenario. Altogether, participating human players can experience and learn the embedded decision makings and complexity as the simulation game proceeds to its full variations.