********************************************************************
ELECTRONIC NEWSLETTER ON REASONING ABOUT ACTIONS AND CHANGE
Issue 99016 Editor: Erik Sandewall 12.6.1999
Back issues available at http://www.etaij.org/rac/
********************************************************************
********* TODAY *********
The refereeing of several articles has been concluded during the last
few weeks, and is reported in today's Newsletter. We have the pleasure
to announce the acceptance of three articles after the standard process
of discussion, revision, and refereeing, namely the article by David
Poole, the reference article by Patrick Doherty et al, and the reference
article by Erik Sandewall. The refereeing for the last one has been
organized in its entirety by Susanne Biundo (who is area editor for
"Planning and Scheduling"). -- The referee reports for these three
articles follow below.
Chitta Baral and Son Cao Tran have made an addition to their article,
the acceptance of which was reported in the previous Newsletter.
The addition was made in response to suggestions by Anonymous Referee 2.
The reviewing of the submitted reference articles has now been
concluded. We hope that authors will make use of the option of
submitting articles where the customary introductory definitions can
be replaced by a reference to the appropriate reference article.
Anonymous Referee 2 for the article by Baral and Cao Tran has
requested that a clarifying addition be made to his referee statement.
It has been installed into the webpage structure.
********* ETAI PUBLICATIONS *********
--- DISCUSSION ABOUT RECEIVED ARTICLES ---
The following debate contributions (questions, answers, or comments)
have been received for articles that have been submitted to the ETAI and
which are presently subject of discussion. To see the full context,
for example, to see the question that a given answer refers to, or to
see the article itself or its summary, please use the web-page version
of this Newsletter.
========================================================
| AUTHOR: David Poole
| TITLE: Decision Theory, the Situation Calculus and
| Conditional Plans
| PAPER: http://www.ep.liu.se/ea/cis/1998/008/
| REVIEW: http://www.ida.liu.se/ext/etai/ra/rac/008/
========================================================
--------------------------------------------------------
| FROM: Anonymous Referee 1
--------------------------------------------------------
REVIEW FORM
===========
1. Are the results of the article, as specified in the summary, of
significant interest for researchers in our area?
The themes considered in the paper are of great interest.
2. Does the full text substantiate the claims made in the summary?
There are many claims, many lacking precision and clarity.
The paper delivers on some of them (see below).
3. Is the text intelligible and concise?
The main ideas in the paper are intelligible, but the presentation
is often confusing (see below).
4. Do you know of any previous publication of the same result?
No.
5. Are there any other considerations that may affect the decision?
No.
MORE DETAILED COMMENTS
======================
Based on my understanding, the best way to describe the paper is as
proposing a way of representing actions, probabilities, utilities,
sensors, and conditional plans in logic programs where:
- actions are represented using a situation calculus formulation close
to Reiter's
- probabilities are represented by treating certain atoms in special way:
the truth of certain combinations of atoms (choices) is set externally,
not by the logic program
- utilities are represented by a "utility" predicate, defined in the
logic program
- sensors are defined by a "sensing" predicate, defined in the logic
program
- conditional plans are defined following the ideas of Levesque
but are interpreted using the "sensing" predicate, rather than
"knowledge" fluents
Once a logic program is defined along these lines and certain conditions
hold, the expected utility of conditional plans is well defined. There
are also hints about how to compute this expected utities by building
suitable "explanations", by this is not explained in detail.
This would be good enough. The main problem is that the author does not
see and does not present the paper in this way.
He sees the logic program in which the truth of certain combinations of
atoms is defined externally as an "independent choice logic", a logic
that is "based on Bayesian decision theory". He also says that the paper
proposes "a new model; not an alternative to POMDPs, Bayesian Networks,
logic programs, and the standard situation calculus, but a way to combine
them", etc.
I think this is quite confusing.
Personally, I'd also prefer less general discussion about decision theory,
logic, uncertainty, etc; and more details on the procedure for
evaluating the expected utilities of plans and its performance over
some examples.
--------------------------------------------------------
| FROM: Anonymous Referee 2
--------------------------------------------------------
Answers to the Refereeing Questions
1. Are the results of the article, as specified in the summary, of
significant interest for researchers in our area?
Yes, definitely.
2. Does the full text substantiate the claims made in the summary?
Yes, fully.
3. Is the text intelligible and concise?
Intelligible: yes. Concise: yes and no; see below.
4. Do you know of any previous publication of the same result?
No.
5. Are there any other considerations that may affect the decision?
None.
General remark. The article is lucid, well written, and easy to
understand. It is relatively long (40 pages), but this is necessary for
covering the topic it sets out for: to propose an integration of the
author's "independent choice logic" with a variety of Toronto situation
calculus, including background, definitions, examples, and comparisons
with other approaches. Everyone may not agree with the approach, but it
is a valuable contribution.
========================================================
| AUTHOR: Chitta Baral and Son Cao Tran
| TITLE: Relating Theories of Actions and Reactive Control
| PAPER: http://www.ep.liu.se/ea/cis/1998/009/
| REVIEW: http://www.ida.liu.se/ext/etaiNIL/rac/011/
========================================================
--------------------------------------------------------
| FROM: The Authors
--------------------------------------------------------
The comments of the anonymous reviewers were very useful. We would
like to especially thank the second reviewer for pointing us towards
the discrete event dynamic system (DEDS) literature and the similarity
between our notion of maintainance and the notion of stability there.
We have now modified our paper by discussing this similarity and
adding a few references to the literature in DEDS.
Chitta and Son
[The authors have provided a summary of the changes, which is available
from the papers review discussion webpage.]
========================================================
| AUTHORS: Patrick Doherty, Joakim Gustafsson, Lars Karlsson,
| and Jonas Kvarnström
| TITLE: TAL: Temporal Action Logics Language
| Specification and Tutorial
| PAPER: http://www.ep.liu.se/ea/cis/1998/015/
| REVIEW: http://www.ida.liu.se/ext/etai/ra/rac/014/
========================================================
--------------------------------------------------------
| FROM: Anonymous Referee 1
--------------------------------------------------------
1. *Does the article represent a tradition or "approach" where there is
already a sufficient volume of work in the field?*
Yes.
2. *Does the article concisely specify the assumptions, motivations,
and notations used in that approach? Does it correctly capture the
assumptions, etc. that have been used and are being used?*
Yes. Note, however, that these notations etc. are fairly complex.
3. *Would reading the present article enable one to skip the
introductory definitions section of many previously published articles
that used the approach?*
Yes. However, the length of the article (32 pages) makes it unlikely
that the same text would ever have been included as a part of another
article.
4. *Is the article also concise in the sense that it does not contain a
lot of material that is unnecessary for the above criteria?*
One would hope that it is sometimes possible to make do with a shorter
introduction to the approach than this one. An article giving a full
account of the approach couldn't have been much shorter, however.
5. *Is the article pedagogical and sufficiently easy to read, but at
the same time precise and correct?*
Yes.
--------------------------------------------------------
| FROM: Anonymous Referee 2
--------------------------------------------------------
I suggest to accept the article. Here is how I feel it conforms
to the various criteria for reference articles:
1. *Does the article represent a tradition or "approach" where there
is already a sufficient volume of work in the field?*
Yes, as is also shown in the article. Moreover it is a very active
approach, with a rapidly growing volume of work.
2. *Does the article concisely specify the assumptions, motivations,
and notations used in that approach? Does it correctly capture the
assumptions, etc. that have been used and are being used?*
Yes concerning assumptions and notations. The article gives a very
clear specification of TAL and in addition a nice "user manual".
Motivations for the types of language statements and an explanation
how the logics deal with the frame/qualification/ramification
problems could be more elaborate: in this respect the article
relies heavily on e.g. "Features and Fluents". However, I suppose it
is acceptable for the article to rely on what the authors call its
companion article (reference [29], on Cognitive Robotics Logic, by
Erik Sandewall).
3. *Would reading the present article enable one to skip the
introductory definitions section of many previously published articles
that used the approach?*
Yes, under the same assumption as above.
4. *Is the article also concise in the sense that it does not contain
a lot of material that is unnecessary for the above criteria?*
Yes.
5. *Is the article pedagogical and sufficiently easy to read, but at
the same time precise and correct?*
Yes.
========================================================
| AUTHOR: Erik Sandewall
| TITLE: Cognitive Robotics Logic and its Metatheory:
| Features and Fluents Revisited
| PAPER: http://www.ep.liu.se/ea/cis/1998/017/
| REVIEW: http://www.ida.liu.se/ext/etai/ra/rac/016/
========================================================
--------------------------------------------------------
| FROM: Anonymous Referee 1
--------------------------------------------------------
I have checked the article against the refereeing criteria for reference
articles:
1. *Does the article represent a tradition or "approach" where there is
already a sufficient volume of work in the field?*
Yes. The article well summarizes research that has been done by
Sandewall and his group over the last ten years.
2. *Does the article concisely specify the assumptions, motivations, and
notations used in that approach? Does it correctly capture the assumptions,
etc. that have been used and are being used?*
Yes. Though clear from the context, I would clearly state the meaning of
the predicate $D([s,t],s)$ used at page 8.
3. *Would reading the present article enable one to skip the introductory
definitions section of many previously published articles that used the
approach?*
Yes.
4. *Is the article also concise in the sense that it does not contain a lot of
material that is unnecessary for the above criteria?*
Yes.
5. *Is the article pedagogical and sufficiently easy to read, but at the
same time precise and correct?*
Yes.
--------------------------------------------------------
| FROM: Anonymous Referee 2
--------------------------------------------------------
General Remarks
The paper is certainly acceptable as a reference article for the features and
fluents approach, and meets all the relevant criteria.
I think it could be improved with the addition of a few examples. In
particular, I think it would be an easier to read if a couple of examples of
action laws were introduced, for standard benchmark scenarios. In fact, I
don't think the form of an action law is ever defined.
It's a shame about the plethora of different ways of writing the same thing.
This is quite confusing for the reader.
********************************************************************
This Newsletter is issued whenever there is new news, and is sent
by automatic E-mail and without charge to a list of subscribers.
To obtain or change a subscription, please send mail to the editor,
erisa@ida.liu.se. Contributions are welcomed to the same address.
Instructions for contributors and other additional information is
found at: http://www.etaij.org/rac/
********************************************************************