[planning & scheduling] the first announcement
Masahiro HORI <HORI@trl.vnet.ibm.com>
Date: Wed, 13 Jan 1993 01:48:29 -0800
Message-id: <9301130941.AA13330@ns3.trl.ibm.com>
Comment: SRKB Distribution List
Originator: srkb-list@isi.edu
Errors-To: neches@ISI.EDU
Reply-To: <HORI@trl.vnet.ibm.com>
Sender: srkb-list@ISI.EDU
Version: 5.5 -- Copyright (c) 1991/92, Anastasios Kotsikonas
From: Masahiro HORI <HORI@trl.vnet.ibm.com>
To: Multiple recipients of list <srkb-list@ISI.EDU>
Subject: [planning & scheduling] the first announcement
<<< SRKB subgroup: knowledge sharing for planning and scheduling >>>
As a special interest subgroup of SRKB phase II activity, we are
looking for ways to facilitate the sharing and reuse of knowledge
used in planning and scheduling. Our long-term goal is to enable
libraries of formally represented, sharable knowledge.
Specifically, we are going to explore representation primitives, which
will be a basis for developing tools that share models and exchange
knowledge at run time. Since scheduling and planning tasks are found in
a variety of application domains such as manufacturing, transportation,
and manning, it should be possible to share models of domains and
problem-solving processes. Thus, our challenge here is to devise a
sharable basis for those knowledge representations, appropriately
abstracting activities in particular applications.
One of the characteristic features of planning and scheduling is
representation of time. Although the ontologies for temporal intervals
and relations are rather generic, they will be a good starting point for
this subgroup's discussion. Another issue will be the resources that
define capacity and capabilities. In the case of production scheduling,
a type of machine provides a certain capacity for processing, and an
order or an operation may consume an amount of that capacity for a
certain period of time. This kind of perspective is relatively common,
but further assumptions and commitments must be clarified with respect
to the details such as measurement units and representation of attributes
associated with resources and consumers, accommodating the differences
among representation systems.
This message is an invitation for you to participate in the activities
of this working group. In accordance with the general strategy of the
SRKB phase II activity, which is to collect, analyze, and experiment
with ontologies, a possible road map of this subgroup's activities
will include the following:
* Introduce ongoing projects related to this topic
- exchange of short descriptions by email
- further exchange of papers
* Discuss similarities and differences in those projects
- with examples of time, resources ...
- clarifying assumptions about the expected scope
* Collect several sample planning and scheduling tasks to be
described by using ontologies
- various testbed examples such as trivial ones, moderate ones,
and real cases.
Of course, the road map is not restricted to the one above, but can
change to accommodate participants' interests. Please note that this
working group activity is not intended to establish a standard for
knowledge representation in planning and scheduling. One of the main
objectives is to capture the rationale underlying the design of related
ontologies. Thus, it is feasible that various ontologies coexist, all
more or less reflecting their own design decisions. This pose an
important research issue that is to accommodate modularized
ontologies, clarifying the assumptions and commitments associated with
those reusable modules. Ultimately, we plan to conduct a small
experiment on translatability among ontologies. Those ontologies may
include ones proposed by other special interest subgroups, since some
of them may be interrelated. The results will be more practical
if subgroups collaborate in appropriate areas.
If you are interested in participating, please contact:
Masahiro Hori <hori@trl.vnet.ibm.com> or
Donald McKay <mckay@vfl.paramax.com>.
CURRENT PROJECTS IN KNOWLEDGE SHARING FOR PLANNING AND SCHEDULING
At least three current projects are associated with this special
interest subgroup. We would like to call for more related projects.
One is the DARPA/Rome Laboratory Planning and Scheduling Initiative (DRPI).
DRPI is currently supporting a group of 30 or so researchers in the fields
of Planning, Scheduling, Knowledge Representation, Case-based Reasoning,
Decision Theory, and Database Management with the objective to build
integrated, cooperative knowledge based systems. The initiative is
organized around a 12 to 18 month cycle of Integrated Feasibility
Demonstrations (IFD) and a 4 to 6 month schedule of Technology Integration
Experiments (TIE). Each IFD must demonstrate a significant contribution
toward solving a critical problem in the domain of Joint Operations
Planning as illustrated by problems posed in planning operations,
communications, and transportation in Desert Shield/Desert Storm. The TIEs
are focused on demonstrating system interoperability at a smaller scale.
Over the past two years, the DRPI Knowledge Representation and Architecture
Issue Working Group has been guiding the development of a knowledge
specification language and system infrastructure to support these
cooperative systems. The Knowledge Representation Specification Language
(KRSL) is the results. KRSL outlines an ontology for representing time,
measurement, objects and concepts, and planning primitives. A shared domain
onotology for Joint Operations Planning is being developed in the the KRSL
language. (For more information on KRSL contact Nancy Lehrer,
nlehrer@isx.com)
The second is the MULTIS project, which has been conducted at Osaka
University since 1987. Major efforts have been devoted to development
of a task analysis interview system for a general class of scheduling
tasks, though the methodology employed is not restricted to scheduling.
In the course of the MULTIS project, a set of two-layered ontologies
were identified for representing problem solving engines for
scheduling and their use in task analysis interview with domain
experts. The higher-level ontology represents human problem solving
process at the knowledge level and is used for the task analysis
interview where it contributes to filling the conceptual gap between
the computer and the domain experts. The lower one consists of a set
of abstract programs used as building blocks of problem solving
engine. The first version of task ontology for scheduling has been
edited. (For more information, please contact Riichiro Mizoguchi
<miz@ei.sanken.osaka-u.ac.jp>)
The third is the CAKE project, which has been conducted since
1990. It was started to establish a methodology for configuring
problem-solving methods with smaller-grained problem-solving components.
A class of scheduling problems is focused on. It is called a job
assignment task, which is characterized as a set of four elements:
jobs, resources, time range, and constraints. The task can be
described as assigning all given jobs to the available resources
within a specific time range, while satisfying various constraints.
Two kinds of ontology are being designed: one is a task ontology for
problem specification; the other is a problem-solving ontology for
representing the problem-solving processes. These ontologies are
defined in KIF and Ontolingua. (For more information contact
Masahiro Hori <hori@trl.vnet.ibm.com>)
-------------