Planning Initiative SPAR - Issue 0006

Date: Fri, 29 Aug 1997 11:21:50 -0400
To: Austin Tate
From: "John F. Lemmer"
Subject: Re: New ARPI Plan Representation Work

My personal feeling is that it may be very hard to design plan representations to provide adequatley for uncertainty. As you may know, Tom Garvey and Doug have asked me to get the ARPI uncertainty and uncertainty research in general focused on the needs of planning. As of now, it is fair to say that _none_ of our uncertainty researchers have any clear vision as to how their research can integrate with the rest of planning on the scale necessary for ARPI. We are currently working on one 'petite' challange problem to see how Markov Decisions Processes (in the large) can support CTEM (in the small) type of planning. There is another 'petite' challenge problem to show how Causal Models can support evaluation of strategy to task plans. This latter problem is farily far along and I think, is ready for demonstration within JFACC.

While evaluation of existing plans in the face of uncertainty is probably ready for demonstration, generation of plans that account, a prior, for uncertainty is in its very basic stages for problems of interest to ARPI. I don't think anyone could accurately forecast the representation needs for this at the current time.

Date: Fri, 29 Aug 1997 16:30:37 +0100
To: "John F. Lemmer"
From: Austin Tate
Subject: Re: New ARPI Plan Representation Work

I think the core meta-models available now ALLOW for uncertainty in some sense, but do not specifically include constructs for it. Maybe this is the best approach given your comments. Hooks and opprotunities are the best that can be offered and maybe the maximum ontological commitment we should make until the uncertainty community itself is more certain of its approach.

Date: Fri, 29 Aug 1997 11:49:23 -0400
To: Austin Tate
From: "John F. Lemmer"
Subject: Re: New ARPI Plan Representation Work

This sounds like a suitable pragmatic approach at this time.

ARPI Home Page | SPAR Home Page

Page maintained by Austin Tate (a.tate@ed.ac.uk), Last updated: Thu Oct 2 18:06:46 1997