Mobile Agents for Networked Electronic Trading
[Summary][Publications][Links]

A mobile agent is a piece of software code that can be sent from one computer to another over the Internet and continue to execute after it has reached the remote site. They are suitable for applications that require a large number of close interactions, and for which communication over the network is costly. An e-commerce transaction requires several interactions between the client site(the buyer) and the server site(the seller). Thus, mobile agents provide a suitable paradigm for modelling e-commerce processes over the Internet. As part of my Ph.D research, I have developed the MAgNET system for comparison shopping over the Internet.

Summary of the MAgNET system

MAgNET(Mobile Agents for Networked Electronic Trading) is a mobile agent based system developed using Java1.1 and the IBM Aglets SDK which enables buyers to comparison shop for items from different online sellers. In MAgNET, a human buyer creates a mobile shopping aglet that compares quotes for an item from different online sellers by visiting those seller sites and returns to the buyer with the best offer that it obtains. The operation of MAgNET can be divided into the following phases:

Buyer Site in MAgNET
Bigger Image
Detailed View 1. A human buyer enters the preferences on different attributes including price, delivery time, and quantity for the item he wishes to purchase, through a GUI on his computer. A stationary controller agent takes this information from the GUI and updates the internal data structures and database on the buyer site. The controller agent then looks up the list of online sellers (stored apriori inside the buyer site's data tables) for each item on the buyer's shopping list. For each item the controller agent creates a mobile shopping agent and gives it information about the item and the list of online seller sites that sell that item.

Seller Site in MAgNET
Bigger Image 2. The mobile shopping agent sequentially visits each seller site on its itinerary. At each seller site it talks to a stationary service agent that interfaces with the database on the seller's backend. The service agent submits the quote from the seller to the shopping agent. The shopping agent compares this quote with the offers from other sellers it has obtained thus far. If the current quote if the best offer yet, the shopping agent reserves the item with the seller and cancels any reservations it might have made at previously visited seller sites.

Complete system view of MAgNET
Bigger Image 3. The mobile shopping agent returns back to the buyer's site at the end of its itinerary of online sellers and submits the URL of the seller that offered the best quote. The buyer can then either contact the seller directly or abort the transaction.

Implementaion

We have simulated the buyer and seller sites of MAgNET by running the Aglets Tahiti server on computers at different locations. The Tahiti server provides a GUI that allows users to create, execute, dispatch and terminate aglets. The seller's inventory is maintained as an XML annotated file that allows browsing both by software agents as well as by humans through a Web browser. The seller's service agent uses the AELFRED XML parser to parse the seller's XML annotated inventory. The data for the sellers' inventory was taken from actual data from different online sellers and stored in the database on the backend of MAgNET's seller sites. We have used the mini-SQL database driven by the m-SQL JDBC driver. The software code for the different aglets was written using Java 1.1.7. The software used in MAgNET are

Java 1.1.6
Java Swing SDK
IBM Aglets SDK 1.0.3
AELFRED XML Parser
Mini SQL 2.0.4
m-SQL JDBC driver

We have also developed a security protocol for MAgNET using the Public Key Infratructure(PKI) that guarantees reliable electronic transactions between buyers and sellers.

Documentation

Snapshots from MAgNET Demo

Item View :The buyer site GUI of MAgNET that enables buyers to select an item by looking at its image and enter preferences of different attirbutes for that item including price, time and quantity.
Progress/Status View :The progress made by the shopping aglet as it completes visiting each seller site on its itinerary. (The nodes in each graph are dynamically updated as the shopping agent leaves a seller site.)
Cart View :The quotes brought back by the shopping aglet for the items on the buyer's shopping list.

[Top]

Publications

P. Dasgupta, N.Narasimhan, L.E. Moser, P.M. Melliar-Smith, "MAgNET: Mobile Agents for Networked Electronic Trading," IEEE Transactions on Knowledge and Data Engineering, Special Issue on Web Technologies, vol. 24, no. 6, July/August 1999, pp 509-525.

P. Dasgupta, N. Narasimhan, L. Moser, P.M. Melliar Smith, "A Supplier-Driven Electronic Marketplace Using Mobile Agents", Proceedings of the First International Conference on Telecom and E-commerce, Nashville,TN, November 19-22, 1998, pp 42-50.

P. Dasgupta, L.E. Moser, and P.M. Melliar-Smith, "The Security Architecture for MAgNET: A mobile agent based e-commerce system," Proceedings of the Thrid International Conference on Telecommunications and E-Commerce, Nov 2000, Dallas, TX

P. Dasgupta, "Fault tolerance in MAgNET: A Mobile Agent E-commerce System," Proceedings of the Sixth International Conference on Internet Computing, Las Vegas, NV, June 24-27, 2002, pp. 733-739

[Top]

Links

Mobile Agent Platforms

IBM Aglets SDK (IBM, Japan)
Jumping Beans (Ad Astra Inc.)

Voyager from Recursion Software (formerly ObjectSpace)

Concordia ( from Mitsubishi Electric)

Odyssey(Inaccessible as on March 1st 2000) (General Magic)

Mole (University of Stuggart)

E-Commerce Applications using Mobile Agents

Links

Cetus Links on MobileAgents

[Top]

[Research Page]
[Home]

Buyer Site in MAgNET Bigger Image Detailed View	1. A human buyer enters the preferences on different attributes including price, delivery time, and quantity for the item he wishes to purchase, through a GUI on his computer. A stationary controller agent takes this information from the GUI and updates the internal data structures and database on the buyer site. The controller agent then looks up the list of online sellers (stored apriori inside the buyer site's data tables) for each item on the buyer's shopping list. For each item the controller agent creates a mobile shopping agent and gives it information about the item and the list of online seller sites that sell that item.
Seller Site in MAgNET Bigger Image	2. The mobile shopping agent sequentially visits each seller site on its itinerary. At each seller site it talks to a stationary service agent that interfaces with the database on the seller's backend. The service agent submits the quote from the seller to the shopping agent. The shopping agent compares this quote with the offers from other sellers it has obtained thus far. If the current quote if the best offer yet, the shopping agent reserves the item with the seller and cancels any reservations it might have made at previously visited seller sites.
Complete system view of MAgNET Bigger Image	3. The mobile shopping agent returns back to the buyer's site at the end of its itinerary of online sellers and submits the URL of the seller that offered the best quote. The buyer can then either contact the seller directly or abort the transaction.