Date: Tue, 10 Dec 1996 15:18:00 GMT
Server: NCSA/1.4.2
Content-type: text/html
Last-modified: Wed, 28 Feb 1996 00:55:40 GMT
Content-length: 3146
Probabilistic Planning
Probabilistic Planning
Adapting classical planning algorithms
to deal with uncertainty about the world state and the effects of
actions:
- The Buridan planner is a first
attempt at a probabilitic planner, and explores issues
like what the action and state representation should
be, and how the problem of evaluating a partially
generated plan interacts with generating a plan.
Buridan assumes the agent will get no additional information
about the world at execution time.
- The Conditional Buridan planner is an extension that allows
actions to provide information about the world as well
as change the world. This is joint work with
Dan Weld
and
Denise Draper.
Recent CBuridan papers include:
- Also see information on the UWL language, which is
a symbolic (non-probabilistic) approach to planning with
information-producing actions and conditionals.
An Approach to Planning with Incomplete Information
by
Etzioni
,
Hanks
,
Weld
,
Draper
,
Lesh, and
Williamson.
hanks@cs.washington.edu