From: Stanley Kok (koks_at_cs.washington.edu)
Date: Fri May 16 2003 - 10:34:00 PDT
Paper Title: Leap Before You Look: Information Gathering in the PUCCINI
Planner
Author: Keith Golden
One Sentence Summary:
This paper shows how partial-order planning can be improved by
adding actions with unknown secondary preconditions to a plan and
verifying the preconditions after the action's execution.
Most Important Idea(s) in the Paper:
1. Secondary preconditions are descriptive and not prescriptive.
Therefore, they can be verified after their associated action is
executed. This removes ordering constraints imposed by the
preconditions and increases the number of admissible plans.
Flaw(s) in the Paper:
1. Since the comparison between VL and ~VL is unfair, the ~VL column
in Table 1 should be removed (since they have no informational value),
and be replaced with the XII columns in Table 1.
2. The paper does not state how many runs of the experiments are
performed. If there is only run, it is unclear how conclusive the
results are since the performances of VL and NO are close. If there
are more runs, the paper could have stated the standard deviations
to better compare the performances.
Important, Open Research Question(s):
1. How can the idea of verifying secondary preconditions after the
execution of their action be extended to probabilistic planning?
2. How can the idea be enmeshed with interleave planning and execution?
This archive was generated by hypermail 2.1.6 : Fri May 16 2003 - 10:35:09 PDT