以內嵌於多位代理人的遺傳演算法求解供應鏈網路平衡模式

Abstract

本研究以多位代理人來建構Nagurney et al.(2003)的供應鏈動態網路，該網路包括位於相同或不同國家的製造商、擔任供應鏈中間者角色的零售商、以及位於不同需求市場以不同匯率貨幣購買產品的顧客，所組成的三階層網路。我們以代理人模式建構代表三階層供應鏈網路成員的實體代理人及位於其中的功能性代理人。透過供應鏈網路供需均衡的實踐，可使整體供應鏈的效益達到最高，並由此獲得供應鏈成員對消費市場需求的產銷分配量。為有效求解供應鏈平衡的問題，本研究將基因演算法內嵌於價格數量均衡的功能性代理人(P-Q)，以取代原論文所採用的變分不等式所推導的條件式及尤拉法，並考量限制條件處理技術，以發展能快速求解且兼具最佳化搜尋能力的基因演算法。在與原論文範例比較的結果顯示，本研究內嵌基因演算法的多位代理人模式可用來建構供應鏈網路模式，而基因演算法可快速求解網路均衡的問題。

We implemented multi-agent for modeling the global supply chain networks studied by Nagurney et al. (2003). The networks contain three tiers of supply chain members: manufactures, who may be located in one or multiple countries; retailers, who act as intermediaries in the supply chains; and consumers at the demand markets who purchase products in different currencies in the countries. The three tiers of supply chain members are modeled as physical agents, while in each physical agent are functional agents. To achieve equilibrium of networks, we used genetic algorithm (GA) embedded in the P-Q functional agent in replace of the original approach of variational inequality formulation and Euler method in solving supply chain network equilibrium. Results showed multi-agent can be a good method in modeling supply chain network optimization problems, and GA performs well in achieving network equilibrium.