Design And Implementation On Agile Supply Chain Management Based On Multi-agent System

Since the beginning of the 21st century, with the rapid development of information technology and diversification of customer needs, the market environment of enterprises changes very much. The market environment dominated by suppliers changes from the past, static, a simple market environment into customer-oriented, dynamic and complex market environment. The competition between enterprises in simplex production quality and performance turns to supply chain of enterprises.Agile supply chain managemen(tASCM) becomes a new mode of operation which adapts to the competitive environment. Although ASCM is developed on the traditional SCM, it has the characteristic which traditional SCM doesn't have, including the property of sensibility to market, web distribution, process integration and virtual property.Agent is the entity which has some properties such as autonomy, adaptation and cooperation, and runs in dynamic environment, it develops from artificial intelligence, and has all kinds of personal attributes, such as faith, intent, expectation, goal, promise, ability, and so on. It can apperceive the environment and react to the environment. Multi-Agent System (MAS) is a dynamic system formed by some related Agents in an open and distributed environment. The Multi-Agent System includes several Agent entities, each Agent has independent function, but they must achieve some task by cooperating with each other.With the deeply studying of Multi-Agent System, the researchers found that there are a lot of common characteristics between the Multi-Agent System and Agile Supply Chain System, So we can make full use of Agent's characteristics such as intelligent, cooperative and adaptive to build a reliable, flexible, and extensible ASCM with Multi-Agent technology.By analyzing the advantage of ASCM based on Multi-Agent,we introduced MAS technology to the architecture of ASCM. We make full use of the advantage of both during design and realization of system, so as to the system becomes more perfect, adaptive and applied.The main content of the paper contains:(1) From the essential characteristics of ASC, we analyze the similar aspects of ASC and Agent System, and design a system architecture of ASCM based on Multi-Agent. The whole model is divided into three layer: user layer, Supply Chain organization layer, and bottom layer namely platform layer. Among them Supply Chain organization layer includes supply firm agent, product firm agent, market firm agent, and CoAgent. Thereinto, CoAgent lies in the top layer of the model and stores the circs of all Agents and information of run status, which is used to negotiate with each other so as to avoid conflict between agents, and negotiate with the information and run of agents in Supply Chain. Each agent module lies in bottom layer and has own state register which is used to store the flags of all movements and the information of run state , and negotiate with the movements of agents. Furthermore, there is exchange of data and information between agents.(2)From the system architecture and design angle, we make a detailed discussion of the core Agents'model structure and their working principle.CoAgent takes charge of layout and management of enterprises , and help lookup and filter when enterprises put forward matching requirement. Then it informs appropriate enterprises and precipitates the negotiation of CoAgent of upriver and downriver enterprises'agents. When enterprises'agents send requirement or data to CoAgents information is received by communication manager and distilled by KQML interpreter, then stored into information base. Reasoning Machine distills knowledge rules from knowledge rules base , deals with data and requirement according to rules and updates rules base. Then changes ultimate result or error information into the visit to corresponding object and answer to enterprises'agents.Product firm Agent may simulate the run and information transfer of product firm which includes local function Agent, shch as run Agent, data management Agent, state management Agent and so on. When Product firm Agent makes a product plan, probably it serves for one node in supply chain, but may pass correlative information to other nodes or receive information from other nodes. When plan needs other information, correlative function Agent of Product firm Agent may inquire. Such as CoAgent may inquire how much will be subscribed to or distill straight information from database; or sale agent may inquire Product firm Agent such information as product and reserve and so on when users want to know when they may receive the goods that they have subscribed to. Thereout we may gain product coordination result which is accurate and timely.Supply Agent includes run Agent, data management Agent, monitor Agent and so on. We design Supply Agent module in order to finish the function of product management system, such as it may simulate the product process of Supply firm, and information exchange between enterprise resource layout system and CoAgent and Product firm Agent. We design sale Agent module in order to manage sale operation,such as sale flow, order form, sale Stat., budget and plan, analysis and forecast and so on.(3)Combining the necessity and traits of supply chain coordination, we design coordination process and communication system between modules in system to all kinds of Agents of ASC that has many Agents. According to the traits of task assignment under supply chain environment we improve contract net protocol and alternation suggestion protocol, i.e. task assignment is realized by the coordination between manager and bidder and between bidder and bidder, and put forward corresponding coordination arithmetic. Based on coordination task assignment , we give effective methods of task assignment under ASC environment and incomplete information and authoritativeless manager.Agile supply chain management's prototype is a chastening experiment system of ASCM system, and validates the feasibility of ASCM system architecture model on technique. With the unceasing research of Multi-Agent and network technique, integrative research of men and machine based on Multi-Agent has wide application foreground to a certainty.