🚀 go-pugleaf

From: XIISJSCH@DDATHD21.bitnet
To: mod-ki@unido.informatik.uni-dortmund.de
Date: Tue Mar 26 14:31:18 MEZ 1991
Subject: <mod-ki 0678> ESWP (report)
Keywords: Berichte

mod-ki message 0678
************************************************************************
                               Report on the
                "European Workshop on Planning (EWSP) 1991"

Submissions to AI-Com, KI, MOD-KI

The "European Workshop on Planning (EWSP) 1991" was held on the  18.-19.
march 1991 at GMD in Birlinghoven (Germany). The workshop was  organized
by Joachim Hertzberg (GMD) and supported  by  GMD  and  GI  (the  german
computer  science  society).  The  castle  of   Birlinghoven   was   the
stimulating  environment  for  17  talks  and  discussions  about   plan
generation and scheduling. About 50 researchers  from  Belgium,  France,
Germany, Italy, Netherlands, Sweden, and UK attended the workshop.

The  intention  of  this  workshop  was  to  bring   together   European
researchers in the field  of  plan  generation  and  scheduling  and  to
familiarize each  other  with  the  work  done  in  Europe.  During  the
workshop, it became clear that such a workshop is really needed  because
there are only national AI-societies with  special  interest  groups  on
planning (UK and Germany).

The papers presented on the two days of the workshop were categorized in
"long" and  "short  papers".  While  the  second  day  focused  more  on
scheduling and the representation and handling of time, the talks on the
first day concentrated on planning and the  (logical)  formalization  of
action.

Long papers:
 ** R. Aylett, A. Fish, S. Bartrum: HELP -- A Hierarchical Execution-Led
    Planner for Robotic Domains
 ** J. Downs, H. Reichgelt: Integrating Classical and Reactive Planning
    within an Architecture for Autonomous Agents
 ** B. Drabble, R. Kirby: Associating AI Planner Entities with an
    Underlying Time Point Network
 ** B. Fronhoefer: Default Connections in a Modal Planning framework
 ** R. Ghassem-Sani, S. Steel: Recursive plans
 ** R. Goebel R. Rodosek: A Language for Representing Planning Problems
 ** G. Grosse R. Waldinger: Towards a Theory of Simultaneous Action
 ** S. Materne, J. Hertzberg: MTMM -- Correcting and Extending Time Map
    Management
 ** P. Regnier, B. Fade. Use of Classical Algorithms for the
    Determination of Parallel Actions and Temporal Optimization in
    Linear Plans of Action
 ** S. Steel: Knowledge Subgoals in Plans

Short papers:
 ** E. Astor, P. Davidsson, B. Ekdahl, R. Gustavsson: Anticipatory
    Planning
 ** E. Babb, H. Barringer, D. Gabbay: HF-PLL -- A Metalevel Rewrite
    Language for Parallel Planning
 ** D. Bernard, M. Borillo, B. Gaume: From Event Calculus to the
    Scheduling Problem. Semantics of Action and Temporal Reasoning in
    Aircraft Maintenance
 ** M. Decker, G. Moerkotte, J. Posegga: Consistency Driven Planning
 ** S. Marcugini, A. Milani, S. Rivoira: Interactive Planning
 ** L. Missiaen, M. Bruynhooge: Localized Abductive Planning
 ** E. Tsang: A Framework for Planning with Incomplete Knowledge
    (this talk was cancelled)
 ** E. Werner: Distributed Planning and Uncertainty

R. Aylett presented HELP, a prototypical planning system for  autonomous
robots in uncertain domains. This system is a  "classical"  hierarchical
non-linear planner based on various levels of world  models.  It  allows
interleaved  planning  and  acting  controlled  by  different   "moods".
B. Ekdahl   introduced   the   concept   of   "anticipatory   planning".
H. Reichgelt described the  integration  of  a  traditional  declarative
deductive planner (TEST) with a  second  planner  using  situated-action
rules (SITACT). To combine the advantages of these techniques, the plans
produced by TEST are automatically compiled into situation-action  rules
used   by   SITACT.   To   provide   greater   flexibility,   hard-wired
situation-action rules can be  integrated  in  this  framework  and  the
planner TEST may be used to pre-compile even more rules.

E. Babb described a planning technique which is  based  on  the  use  of
rewrite rules to manipulate goals, situations, and plans. The underlying
logic integrates several sorts  including  intervals  and  time  points.
S. Marcugini  showed  how  interaction  with  the  environment  can   be
integrated in the planning procedure, in  order  to  solve  "impossible"
goals. S. Steel demonstrated, how recursive plans can be  created  based
on  the  induction  principle  of  Boyer  and  Moore.  The  approach  to
"consistency driven planning" was presented by  G. Moerkotte.  He  views
planning as the manipulation of a database under  special  consideration
of consistency constraints. The  idea  of  linear  proofs  was  used  by
B. Fronhoefer to generate linear plans.  He  extends  this  approach  in
various ways and discusses its properties. In  the  final  talk  of  the
first day, G. Grosse proposed a logical  formalization  of  simultaneous
actions and applied a deductive synthesis method to generate  respective
plans.

The second day of EWSP started with  talks  of  S. Steel  and  E. Werner
presenting   modal   characterizations   of   the   agents    knowledge,
uncertainties, and temporal relationships in the planning  domain.  Very
similar, the cooperation of multiple  agents  and  their  abilities  are
handled as modalities. P. Regnier presented a method -- based on  STRIPS
triangular tables -- which allows to determine which actions of  a  plan
can be  executed  in  parallel.  B. Drabble  gave  an  overview  of  the
architecture of the O-PLAN2 system. The  goal  of  this  project  is  to
handle large planning and scheduling problems with thousands  of  action
nodes in  industrial  environments.  Therefore,  the  system  integrates
traditional  techniques  for  handling  plans,  goal   structures,   and
constraints with a time  point  network  (TPN).  This  seems  necessary,
because the overall efficiency of a  planner  depends  heavily  on  fast
computations of time. The talk of J. Hertzberg gave a discussion of  the
correct clipping of time intervals in the  time  map  management  system
MTMM.

D. Bernard described  the  application  of  the  event  calculus  to  an
aircraft maintenance domain and showed how this approach can be extended
to  support  plan  generation.  R. Goebel  derived  requirements  for  a
scheduling system and developed on this basis a  language  to  represent
temporal constraints. In the final talk  of  the  workshop,  L. Missiaen
described a planner named CHICA which combined abductive  reasoning  and
the event calculus in a logic programming framework.

During lunch time, there were computer  demonstrations  of  the  systems
CHICA and HELP.

Like at every other workshop about planning, the  discussion  about  the
blocksworld and "real"  problems  was  not  missing  at  the  EWSP.  The
speakers began to classify their talks whether blocksworld examples  are
used or not. But in general, the  statement  of  J. Hertzberg  that  the
blocksworld is a "didactic example domain" was widely accepted.

At the closing discussion and during the conference dinner (in  Bonn  at
the end of the first day), the subject of the debate was whether such  a
workshop should be repeated and if so,  which  frequency  would  be  the
best. A main problem seems to be, that there are quite a lot of  related
workshops and conferences. It was considered to integrate  the  workshop
with  ECAI.  However,  the  idea  to  hold  the  EWSP  every  two  years
alternating with ECAI seemed to attract more sympathy.

At the end  of  the  workshop,  it  was  very  likely  that  proceedings
(including the long papers) will be  published  somewere.  Requests  for
information about EWSP'91 and comments on  the  future  organization  of
similar workshops should be directed to Joachim Hertzberg  (address: GMD
(F3), Postfach 1240, 5205 Sankt Augustin 1;  email: hertz@gmdzi.gmd.de).

Josef Schneeberger
FG Intellektik, FB Informatik,
Technische Hochschule Darmstadt,
Alexanderstr. 24, D--6100 Darmstadt (Germany),
Phone: +49 (6151) 16-5469, email: xiisjsch@ddathd21.bitnet
************************************************************************