Planning with Loops

Hector Levesque

University of Toronto

Automatically generating plans of action remains a central problem in AI research. Unlike the case for sequential and conditional planning, much of the work on iterative planning (planning where loops may be needed) leans heavily on theorem-proving. In this talk we do the following: propose a different approach where generating plans is decoupled from verifying them; describe the implementation of an iterative planner based on the language of the situation calculus; present a few examples illustrating the sorts of plans that can be generated; show some of the strengths and weaknesses of the approach; and finally sketch the beginnings of a theory, where validation of plans is done offline.

Bio: Hector Levesque received his Ph.D. from the University
of Toronto in 1981.  After graduation, he accepted a position at the
Fairchild Lab for AI Research in Palo Alto, and then joined the
faculty at the University of Toronto where he has remained since 1984.
Dr. Levesque has published over 60 research papers, and is the
co-author of a recent textbook on knowledge representation and
reasoning.  In 1985, he received the Computers and Thought Award given
by IJCAI.  He is a founding fellow of the AAAI, and was a co-founder
of the International Conference on Principles of Knowledge
Representation and Reasoning.  In 2001, Dr.  Levesque was the
Conference Chair of IJCAI-01, and served as President of the Board of
Trustees of IJCAI from 2001 to 2003.

Deepak Ramachandran

Last modified: Wed Dec 7 19:16:52 CST 2005