From: Stanley Kok (koks_at_cs.washington.edu)
Date: Tue May 20 2003 - 10:20:44 PDT
Paper Title: A Knowledge-Based Approach to Planning with Incomplete
Information and Sensing
Authors: Ronald P. A. Petrick, Fahiem Bacchus
One-line summary:
This paper presents a knowledge-based approach to planning.
Most Important Ideas in the Paper:
1. By formulating the agent's knowledge as a knowledge base, a
planner can abstract irrelevant distinctions that occur in possible
worlds, thereby preventing the number of possible worlds from growing
exponentially (with the number of propositions). In addition, the
planner has a rich representation capable of modelling functions
and runtime variables.
Flaw in the Paper:
1. Even though the paper states that its planner's performance is
comparable to those of other planners such as HSCP, CMBP, and GPT,
it does not provide experimental comparisons (it only
shows the runtimes of its PKS). The reader would have appreciated
a comparison of how fast PKS is relative to other planners.
Important, open research questions:
1. How can the knowledge-based approach be extended to handle
parallel plans?
2. What heuristic search can be used to replace the blind search in
PKS?
This archive was generated by hypermail 2.1.6 : Tue May 20 2003 - 10:21:54 PDT