| Agents for online trading purposes can be seen as tools that help Internet users to purchase products from distributed resources based on their interests and preferences. One of the major features that determine the success of a trading agent is the ability to negotiate with other agents, because most trading tasks involve interaction among agents. This thesis presents a peer-to-peer multi-agent system architecture for online trading. The main objective of this system is to address some of the shortcomings that are present in contemporary online trading systems, such as single attribute-based negotiation, the requirement of an electronic marketplace and variations and status changes within the network. The proposed system architecture is a multi-tier, multi-agent architecture. The agents of the system are autonomous, goal driven agents that work together cooperatively to assist different users to buy products online based on their interests and preferences. The system architecture consists of three types of agents that are classified based on their functionality: interface, resource and retrieval agents. The interface agents are the front-end of the system and able to interact with different users to fulfill their needs. At the middle-tier, the resource agents access and capture the content and the changes of the local information database. The retrieval agents are the back-end of the system and able to travel and interact with other agents at remote host machines. A prototype that employs the system architecture is implemented using the IBM Aglet SDK. |
