Prodigy /Agent Home Page

PRODIGY is a planning and learning architecture developed at Carnegie Mellon University. The core of the system is Prodigy 4.0, a nonlinear state-space planner. We have implemented a wrapper in Allegro Common LISP 6.0 to "agentify" PRODIGY. In the resulting Prodigy/Agent system, a simple client-server relationship exists for interaction. Prodigy/Agent is the server and the human or another system is the client. Our conversation protocol is implemented in KQML with sockets.
 
 

Research Goals

• Enable PRODIGY to collaborate with other agents.
• Develop a generic interface for PRODIGY to allow use by a variety of different applications.
• Enable concurrent instantiations of PRODIGY to collaborate on a common task.

Downloads

Customizing Plan Format

Defined Conversations

Publications

Sample Interaction with Application

Future Research Goals

• Make the Prodigy/Agent package publicly available.
• Define more flexible conversations.
• Study the costs/benefits of using multiple instantiations of PRODIGY to work on a single problem.
• Evaluate Sidner's protocol as an alternative to KQML.