《A FIPA-OS Based Multi-Agent Architecture for Global Supply-Chain Applications.doc》由会员分享，可在线阅读，更多相关《A FIPA-OS Based Multi-Agent Architecture for Global Supply-Chain Applications.doc（6页珍藏版）》请在三一文库上搜索。
1、A FIPA-OS BASED MULTI-AGENT ARCHITECTURE FOR GLOBAL SUPPLY CHAIN APPLICATIONSMihaela Ulieru, Mircea Cobzaru and Douglas NorrieIntelligent Systems GroupThe University of Calgary, Alberta, Canadahttp:/isg.enme.ucalgary.caABSTRACTBased on our Multi-Resolution Collaborative Architecture (MRCA) 1 develop
2、ed for global manufacturing applications, we propose a novel solution to manage global supply-chains to implement MRCA using the latest standards of the Foundation for Intelligent Physical Agents (FIPA) http:/ Methods are given to implement communication-cooperation layers and functional levels of M
3、RCA are proposed to identify suitable FIPA entities and practical solutions to implement global supply chain applications. Our vision for extending the FIPA-OS architectural elements to develop and implement generic web-Centric collaborative applications concludes the paper.IntroductionRecent advanc
4、es in information technology have dramatically challenged how enterprises do business by replacing traditional off-line information processing methods with on-line real-time connections to the global economy. From e-commerce and marketing intelligence to custom-based product design and ordering, mor
5、e and more transactions flow over the internet making distances between business partners measurable in “network intelligence power” rather than in thousands of miles. The increasing flexibility in choosing trading partners has opened ways to partnerships which were unthinkable before. Enterprises c
6、an constantly choose the best suppliers and customers can access the best deals in real time keeping up with the fast paced environment of our current economy. Strategic partnerships are dynamically formed based on the latest information on “whats out there” and virtual enterprises emerge from clust
7、ers of best partners in the global supply chain. On-line inter-enterprise collaboration is the most powerful way to keep up and lead in todays global economy.For 10 years, researchers of the Intelligent Manufacturing Systems Group at the University of Calgary have been developing multi-agent systems
8、 for manufacturing applications 1-5. The essence of this research has been encapsulated into a versatile and generic recursive multi-resolution collaborative architecture (MRCA) that supports web-centric global manufacturing from the high level of inter-enterprise collaboration in the supply chain d
9、own to the low production control level 6, Fig. 1. Fig. 1. MRCA: Communication-Collaboration Perspective.A detailed description of this architecture is presented in 6 here we will focus only on the implementations details using the recent standards developed by the Foundation for Intelligent Physica
10、l Agents, namely the FIPA-OS standard 7.Main Architectural Elements of the FIPA-OS StandardThe FIPA-OS architecture provides an agent software framework and brings interoperability within and across agent-based applications. It employs intelligent agents with implementations based on Java and Intern
11、et-relevant technologies (XML, SMTP, Jini, Active Objects) - elements which have already become regular ingredients in agent applications. The interoperability involves relationships between agents, between agents and platforms and relationships between implementations of agent services. The FIPA OS
12、 architecture in Fig. 2, proposes a concept of an Agent Platform (AP) to be built on top of a distributed computing environment, thus integrating the client/server paradigm with agent technology.Fig. 2. Main Elements of the FIPA-OS Architecture.As depicted in Fig. 2, the AP consists of three basic s
13、ervices: the Agent Management System (AMS), the Directory Facilitator (DF) and the Agent Communication Channel (ACC). When agents are registered with the AMS for a platform they are considered being part of that platform. The AMS is responsible for the management of operations on the agent platform
14、as well as for the management of the agents themselves. The DF (an agent itself) works in a yellow pages manner and supports the localization of agents and the services they provide in the area of a domain or the whole environment. In order for the agents to be registered with a DF, they have to sen
15、d a registration request to that service. The ACC uses information provided by AMS to enable agent communication between agents on a platform and between platforms by offering a message forwarding service. The high level of interoperability between platforms is made possible by placing a mandatory A
16、CC agent on each agent system belonging to a FIPA compliant environment.To support IIOP (Internet Inter-Orb Protocol), which is the necessary protocol for interoperability, the ACC agent is enabled with a CORBA interface. Thus, as a CORBA object, the ACC agent is provided with methods that can be in
17、voked via IIOP. The communication between FIPA agents takes place through the specific communication language, FIPA-ACL.The IORs (Interoperable Object References) must be distributed among the FIPA platforms so that interoperability can be achieved. Even though the IORs are not humanly readable, the
18、y are strings that can be transmitted in many ways: ftp, www, SMTP. Different objects and/or programs can interact with one another through an ORB (Object Request Broker). Therefore, agents and applications, will each have their own brokers, which can carry on the message exchange through the IIOP s
19、pecifications, for that scope using the TCP/IP layer that provides the port numbers and hostnames for reliable delivery. In other words, the ORB layer is responsible for directing and translating the IIOP requests. With current implementations of ORBs, it is not possible to use the URL as an address
20、 for a platform; hence, the solution is to share a directory in which the platforms will have their own files containing the corresponding IORs.The Domain in Fig. 2, is a virtual space (area) that facilitates the management of different application entities (supplier, order, search, transport, etc.)
21、 and encapsulates certain capabilities and restrictions for residing agents providing uniformity in the interactions that take place between software modules.Implementing MRCA for Supply-Chain Applications using the FIPA-OS StandardGeneric Architectural ElementsThe functional perspective on MRCA 6,
22、Fig. 3, consists of an agent hierarchy (vertical) and a communication platform hierarchy (horizontal). The Communication-Cooperation layer in MRCA (Fig. 1) is implemented via several FIPA-OS Agent Platforms (Fig. 2) that support specific services as follows: Management and registration (AMS in Fig.2
23、 which implements the Messaging Management level in MRCA, Fig. 3); Communication (ACC in Fig. 2 which implements the Communication Services Management level in Fig. 3); Agent Interaction (Domain in Fig. 2 which implements the Conversation Management level in Fig. 3) ensuring uniformity in the intera
24、ctions that take place between software modules; Yellow Page Services (DF in Fig. 2 which implements the Collaboration Agents level in Fig. 3);Fig. 3. MRCA Functional Perspective.Besides these generic architectural features, the implementation of global supply chain applications requires specific pr
25、ocedures and control elements. Supply-chain Application-Specific Agents, Fig. 2, implement the Domain Agents (i.e. roles such as inventory control, material flows, information storage, etc.) and the Resource Agents (i.e. capacity allocation, resource management, information management, etc.) in Fig.
26、 3. Specialized Agents for Supply Chain ManagementApplication specific agents (Domain and Resource Agents in MRCA, Fig. 3) needed for supply chain implementation are shown in Fig. 4 as being served by Information Agents (Knowledge Management Agents in MRCA, Fig. 3). The Information Agents (Fig. 2) a
27、re specific implementations of the Knowledge Management Agents (Fig. 1,3) which are used to manage information flow between the various Resource and Domain Agents. Their ability for flexible interaction and interoperation with other software agents makes them emerge into an infrastructure - an infor
28、mation inter-space where information is stored, maintained, converted and shared among specialized agents.Fig. 4. Supply-Chain Application Specific Agents and their InteractionsInformation agents also carry out communications with functional agents mediating message exchanges among them, routing and
29、 distributing at the same time the required information in the right form to those that need it. The specialized agents that perform supply chain functions can store information and knowledge locally and also access information and knowledge throughout the network using services mediated by Informat
30、ion Agents. The Supply-Chain Application Domain Agents are: Shipping Agents are responsible for allocating and scheduling transportation resources required by operation planning agents. Among their data, we can find different transportation routes, schedules and the associated costs; Order Agents ar
31、e responsible for handling requests from customers, approving orders and getting information regarding the orders. This information is then passed to the operation planning agents. Operation Planning Agents detail customer order fulfillment and handle distribution of materials and products across th
32、e supply chain network. Legacy applications and Production Agents are responsible for operations that involve planning, scheduling and manufacturing of products based on demands. Resource Agents are production-specific agents that evaluate the network capacity by comparing the capabilities with the
33、order requirements and send feedback to the Operation Planning Agents regarding possible schedules. They manage resource availability related to suppliers, production and transportation to maximize efficiency so that a certain schedule can be accomplished. The Databases contain information about inv
34、entory-products that are already in stock which were built based on demand forecast, or parts and materials sorted by name, cost and quantity required to manufacture a certain product.FIPA-OS based MRCA for Global Supply Chain ManagementImplementation of the generic Multi-Resolution Collaborative Ar
35、chitecture (Fig.1, Fig. 3) using FIPA-OS architectural standards (Fig. 2) is shown in Fig. 5. The Protocol and Communication Layer in Fig. 5 implements the Communication-Collaboration Layer of MRCA (Fig. 1) using FIPA specific standards, as described previously (see Fig. 2). The SCM Specialized Agen
36、ts are built using a template and then plugged into the Agent Layer as Resource or Application Domain Agents. The Platform Agents manage local inter-agent communication as well as inter-platform communication over the distributed application. The architecture in Fig. 5 supports customer-centric glob
37、al supply-chain applications and can fulfill any order on-line while optimizing costs and delivery time by taking advantage of the enabling internetworking technologies. Once an order for a specific product has been placed by a customer, the inter-agent communication-cooperation for order fulfillmen
38、t (Fig. 4) takes place as follows. The Order Agent contacts the Information Agent who has knowledge about the products design mapped into a database of potential collaborators in the production and/or delivery of the different parts and/or materials. According to the priorities, all collaborators ar
39、e contacted via the inter-platform communication supported by the ACC/FIPA-ACL mechanism and those which commit to deliver in the shortest time are selected. For each partner, an Agent Platform is in place, as in Fig. 2. The detailed description of the inter-agent collaboration in a case study invol
40、ving a multi-national corporation has been presented previously 6. In this paper, we focus on the FIPA-OS mechanisms that facilitate its implementation. The selected partners for the parts and/or materials will register with the main manufacturer and proceed to plan their individual jobs using the O
41、peration Planning Agents that will dispatch the operations on each resource involved in production. The control mechanism for production control at the factory level is detailed in 6.Fig. 5: Supply-Chain Specific MRCA Implementation according to FIPA-OS Standards For each partner involved in the col
42、laborative application, the Supply Agents provide all the necessary materials and supplies for the specific order to be available in due time. The Shipping Agents manage delivery in due time to the main manufacturer of all the parts, once they are ready. Here the final product is assembled and deliv
43、ered via its domain-specific Shipping Agents to the customer.FIPA-OS as an Enabling Technology for Global Inter-Enterprise CollaborationThe mechanisms that facilitate global supply chain management can be extended to any kind of global inter-enterprise collaborative application requiring web-centric
44、 e-business concepts. Using optimization methods to cluster the most appropriate partners for a specific collaboration, we have developed emergent organizations mechanisms capable of rapid response to the ever-changing demands of a fast-paced economy 8, 9. These mechanisms play an important role in
45、the implementation of global organizations which base the essence of their existence on information exchange between several organizational entities distributed at various locations around the globe. The combination of multi-agent technologies with the web-centric paradigm opens wide perspectives to
46、wards development of mechanisms to create virtual organizations with a flexible, self-organizing structure. Unfortunately the current FIPA specifications do not address the implementation of web-centric paradigm specific elements but can be easily extended to cover e-business aspects by building age
47、nts to perform order-taking, accounting and security tasks. Moreover, software agents can collect information on products and suppliers that can fulfill the orders, evaluate biddings, make decisions on them, negotiate terms of transactions, place orders and make payments to the parts involved in the
48、 supply chain network. To integrate these technologies seamlessly, there is a need to define universal standards for goods and services, consumer and vendor profiles, nature of services, secure payment mechanisms, and business electronic forms. Issues such as security, performance, and failure may b
49、e handled by proper design, while others such as communication and interoperability between applications can be managed using so-called middleware solutions (CORBA, DCOM), additional software platforms (Voyager) or built-in language mechanisms (Java RMI) and web-specific meta-languages (XML). There is also a trend to integrate pre