Abstract: This paper proposes an agent-based grid architecture to
support case-based reasoning on grid (CBRG) to expand users’ experiences
sharing globally. This architecture comprises four parts: User Agent, Task
Agent, Directory and Monitoring Agent (DM Agent), and the foundation grid
layer. The three agents use the services provided by the grid layer
to communicate with each other and cooperate to finish the overall CBRG process.
The resources involved in CBRG are categorized into three classes: problem,
case and case library, and implementation software. Each class is modeled
by a metadata document formulated using XML. Based on the metadata model,
the interactive mechanism and processes among the three agents are explored.
The functions and structure of each agent are also analyzed. The User
Agent intelligently guides user through case representation and other interactive
processes. The Task Agent generates execution plan and schedule the implementation
process which supports distributed parallel reasoning. The DM Agent registers
resources metadata, monitors the change in resources and triggers the optimization
process to enhance the opportunities of getting a better result for user.
Through mining the data stored in the metadata library and execution library,
the DM Agent can find the user preferences and cluster users into user groups
while these information can greatly facilitate the reasoning speed and quality.