《毕业设计论文 外文文献翻译 电子商务专业 引入第三方物流企业的以代理人为基础的电子商务系统模型 中英文对照.doc》由会员分享，可在线阅读，更多相关《毕业设计论文 外文文献翻译 电子商务专业 引入第三方物流企业的以代理人为基础的电子商务系统模型 中英文对照.doc（16页珍藏版）》请在三一办公上搜索。
1、2007 IEEE/WIC/ACM International Conference on Intelligent Agent TechnalogyIntroducing Commodity Flow to an Agent-Based Model E-commerce SystemTomasz SerzyskoDepartment of Mathematics and Information Science,Warsaw University of Technology,PolandMaria Ganzha,Maciej Gawinecki,Pawel Kobzdej,Marcin Papr
2、zyckiSystem Research Institute,Polish Academy of Sciences,PolandCostin BadicaDepartment of Computer Science,University of Craiova,RomaniaAbstractIn our model agent-based e-commerce system 2 we have assumed that a certain number of items of a given product is available for sale. In this note we intro
3、duce a model logistics subsystem and discuss how it will be integrated with the system.Keywords: e-commerce;logistics system;agent1. IntroductionCurrently, we are developing and implementing a model agent-based e-commerce system (see 2 and references collected there). In this system multiple buyer a
4、gents attempt at making purchase by participating in price negotiations in e-stores and selecting the best oer, while e-stores attempt at maximizing profit resulting from product sales.Thus far our attention was focused on buyer-seller interactions. By assuming that products are in the warehouse we
5、have omitted the question where do they come from. The aim of this work is to describe how our system can be extended to include product restocking processes.Before proceeding let us make a few observations. First, the proposed logistics subsystem is not “stand alone” (e.g. similar to that considere
6、d in 3, 1, 6). Instead, it has been created within the context of our e-commerce system, which has directly influenced its design. Second, while somewhat similar, processes involved in e-store restocking a warehouse differ from client making a purchase in an e-store (e.g. in product demand predictio
7、n, interactions with wholesalers, methods of price negotiations that involve more “conditions,” offer selection criteria, etc.) Therefore we have created a separate logistics subsystem (instead of reusing already modeled functions; e.g. price negotiations). Third, this note is devoted to the agent s
8、tructure and agent interactions and, due to the space limitations, we omit important topics like:forecasts derivation, oer evaluation etc.However, these functions can be encapsulated into modules that can become a part of an appropriate agent. Therefore readers should envision that, for instance, wh
9、en we write that “received oers are evaluated,”then their favorite method of oer evaluation has been utilized.To proceed, first, we briefly describe our e-commerce system.We follow with assumptions that underline the logistics subsystem and description of new agents that were introduced. Finally, we
10、 present the sequence diagram of restocking and use it to discuss in detail how this process will take place in our system.2. System DescriptionOur system is a distributed marketplace in which software agents perform e-commerce functions (see 2 for details, the Use Case diagram in particular). User-
11、Client is represented by the Client Agent (CA). The CA is autonomous and when a purchase order is communicated by the User-Client, it works until either it is completed, or purchase is abandoned. The CA communicates with the Client Information Center (CIC), which facilitates information which e-stor
12、es sell which products (yellow-page matchmaking).For each store that sells the requested product, the CA delegates a Buyer Agent (BA) to participate in price negotiations and if successful, possibly attempt at making a pur- chase (successful price negotiations result in a product reservation for a s
13、pecific time; after which products that have not been purchased are available for sale again).Since multiple BAs representing the same CA can win price negotiations the CA makes the decision if either of available oers is good enough to make a purchase. Buyer Agents can participate in negotiations o
14、nly if the Gatekeeper Agent (GA) admits them (if they are trusted; e.g. BAs that win price negotiations but do not make a purchase may be barred from subsequent negotiations). The GA represents a given e-store and is created by the Shop Agent (SA). The SA is the central manager and facilitating the
15、selling process it utilizes the GA, and a set of Seller Agents (SeA) that negotiate price with incoming BAs, as well as a Warehouse Agent (WA) that is responsible for inventory and reservation management. Thus far, the WA was responsible for managing product reservations and the inventory. Specifica
16、lly, (1) before a new price negotiation the WA “reserved” a given productso that if negotiation ended successfully there was an item that could be sold; (2) when a reservation ended in purchase, it adjusted product counters; and (3) when a reservation expired it also adjusted products counters. Howe
17、ver, the WA was always envisioned as the “gateway” between the store and suppliers, which is one of the foci of this note.3. Assumptions behind the logistics systemLet us now specify assumptions that underline design of the logistics subsystem: 1.Nowadays, except of largest store-chains (e.g. TESCO,
18、 WalMart), companies outsource transportation activities (considered non-core business activities) to specialists (e.g. UPS, Mayflower). However, we assume here that suppliers are still responsible for facilitating transportation. Therefore, we omit transportation related processes and focus only in
19、 interactions between e-stores and suppliers2. As a result of (1), transportation costs are assumed to be paid by the supplier and included directly in product price (e.g. discount on delivery costs, will manifest itself in the total price).3. While in the original system auction-based price negotia
20、tions were used, here we opted for the simplicity of the FIPA Contract Net Protocol 4. Therefore, in the logistics subsystem, a single round of negotiations consisting of a call for proposals and evaluation of responses, is used.4. New functions and agentsIn order to perform logistics-related tasks,
21、 several new roles were introduced; some of them have been delegated to agents existing in the system, while others warranted adding new agents. Specifically: demand estimationto draw information from sales data and/or external premises to predict future sales of products, warehouse monitoringto obs
22、erve supply levels and react in case of a risk of dropping below values considered sucient to satisfy estimated demand, order managementto coordinate issuing orders for goods, to assist in evaluating received oers, ordering goodsto contact suppliers for their of- fers and to select the best oer, sel
23、ling goodsin the system suppliers were also modeled; while goods are acquired without extensive price negotiations, “someone” has to deliver proposals to ordering components, logistics yellow pagesthe role of the “logistics CIC” is very similar to the original CIC (2); it has to provide lists of pot
24、ential suppliers of products; obviously, it is possible for a shop to become a sup- plier for another shop and to suggest that the two CICs could be joined, but we decided to clearly separate the client-side from the shop and from the supply-side. Another reason for this separation was that while so
25、me shops may not be interested in becoming wholesaler, we would have to make changes to the original CIC data structure (e.g. wholesaleryes/no). Finally, since the logistics subsystem does not involve auctions, the separation is even more warranted.Let us now see how these tasks/roles could be place
26、d in our system. The demand estimation role was attributed to the existing Shop Decision Agent (SDA), responsible for the “knowledge management” functions (e.g. trust management, sales trend data mining, etc.) in the shop.The warehouse monitoring role is already a part of the existing WA. The dieren
27、ce is that now WA becomes a proactive manager of supplies; acting on predictions supplied by the SDA.Fulfillment of the order management role required introduction of the Logistics Agent (LA), which became the “central manager” of the logistics subsystem. It is responsible for contacting the logisti
28、cs CIC for the list of potential suppliers and managing a pool of agents responsible for ordering goods from “wholesalers.” Finally, it collects and manages data related to supplier reliability. This data, in turn, will be one of factors in selecting the supplier.The ordering goods process is facili
29、tated by the Ordering Agent (OA), which is also a new agent. Its task is to issue a call for proposals, collect responses and select the best oer taking into account factors such as: price, delivery time, reliability etc. Let us recall that due to the modularity of agent design (2), our system is fl
30、exible enough so that any method of selecting an oer can be applied (it can be encapsulated in a module and plugged into the OA).The selling goods role is realized by a very simple Wholesale Agent (WhA). Its role is to respond to CFPs incoming from OAs.Currently we assume that WhAs receive instructi
31、ons in what way to generate a stream of responses to the CFPs.Finally, logistics yellow pages are facilitated by the logistics CIC Agent. Its role is to store a complete list of suppliers and products that they sell. Obviously, the logistics CIC uses the original product ontology (2), extended by th
32、e logistics ontology. When the system is initialized, each WhA registers with the logistics CIC and provides it with a list of products for sale.What was described thus far is summarized in an UML use case diagram presented in Figure 1.Figure 1. Use Case of the logistics subsystem5. Typical Product
33、Restocking ProcessLet us now describe the processes involved in restocking the warehouse. Here, we skip the description of system initialization, and start with the Shop Decision Agent sending a forecast to the Warehouse Agent. The sequence of actions resulting form such a forecast is depicted, as a
34、 UML sequence diagram in Figure 2.The SDA communicates the forecast to the WA by sending a FIPA Inform message containing the PredictionDescription (which contains all necessary data such as: product ID, amount of predicted sales, standard deviation of sales, expected purchase price, period for whic
35、h this forecast is valid, etc.). We assume that the SDA forecasts are of the type: until a new fore- cast, weekly sales are expected to be 45 items of a given product. Forecasts can be issued at specific times (e.g. once a week or once a month) and their frequency de- pends on the information found
36、in data analyzed by the SDA to derive forecast(s).The WA starts by examining current stock of a given product, and if current supplies are sucient, it sets up to check their levels at the end of the time unit specified in the forecast (i.e. forecasts specified on weekly basis are checked once a week
37、). If stocks are insucient, the WA utilizes the FIPA Request Protocol (FIPA specification SC00026) and FIPA SL language 4 (used in all agent interactions), to com- municate with the Logistics Agent. The initial message from the WA is the FIPA Request message sent to the LA and it contains OrderReque
38、st action with the Or- derRequestDescription. The OrderRequestDescription contains the necessary information specifying the order to be made: product ID, preferred delivery time, amount and maximum price. Delivery time and amount are computed based on the current product level, predicted delivery ti
39、me and an overall inventory strategy.Upon receiving the request, the LA dispatches a query to the logistics CIC to obtain a list of suppliers of a given product. Ensuing conversation conforms to the FIPA Query Protocol, starting with the FIPA Query-Ref message containing the CICQuery action with the
40、 Product ID. The logistics CIC responds with the FIPA Inform-Ref message containing the CICResponse with a list (possibly empty) of suppliers. Empty list results in a FIPA Failure message (with OrderRequestResult set to failure) send by the LA to the WA. Similar response is sent when the logistics C
41、IC cannot be contacted.When the non-empty list was received, the LA removes these suppliers that have their reliability value below a certain threshold. Then the LAAgentDescriptions list is formed by supplementing each CICAgentDescription received from the logistics CIC with the reliability informat
42、ion. If a given WhAs is not known a default trust value is used.After preparing the list, the LA utilizes the FIPA Request Protocol to find a free OA. Busy agents will re- spond with FIPA Refuse messages. If all agents respond in such a way, this process may need to be repeated un- til a free OA is
43、found and responds with the FIPA Agree message. The LA then sends the ordering request to the selected OA and awaits for the result of the ordering pro- cess. LAs message contains the IssueOrder action with the OrderDescription and the LAAgentDescriptions.After obtaining the request from the LA, the
44、 OA engages in the FIPA ContractNet Protocol interactions with WhAs from the list. It sends the FIPA CallForProposal message, containing CFPRequest with OrderDescription to the WhAs. WhAs evaluate the CFP and submit their oers by sending FIPA Propose messages containing the CFP Response action with
45、OerDescription or, if terms contained in the CFP are unacceptable/not interesting, respond using the FIPA Refuse message.Responses must arrive within a timeframe speci- fied by the OA, after which the OA proceeds to evalu- ate them. First, it filters unacceptable oers. Note that it is possible that
46、some WhAs may respond knowingly with proposals that violate some of the conditions and in special circumstances when no better oers were found the OA may need to accept such oers. In the next step, oers are ranked and the winner is determined. Winner is sent a FIPA AcceptProposal message containing
47、the ConfirmationRequest action with its offer quoted. The winning WhA must in turn reply with the FIPA Inform message containing Confirmation Response action with the OrderConfirmation which has unique orderID generated by the supplier. This successfully completes the ordering process. The winner ca
48、n also withdraw the oer by sending a FIPA Failure message. In this case, runner-ups are contacted in an iterative manner. In case when there are no more oers left or there were no oers to begin with, the OA sends a FIPA Failure message to the LA, which, in turn, forwards it to the WA. When the winne
49、r confirms the order, the OA sends to the LA a FIPA Inform message containing the InformResult action with the WhA-received Order Confirmation, thus completing the protocol. At this time the LA sends information to the WA, inside a message of the FIPA Agree type. This performative is used in compliance with the protocol to indicate that the LA is performin
