Planning Representation Input to JTF ATD C2 Schema

Austin Tate (, 3-Jul-96
This page acts as a reference source for the July 1996 Workshop on the JTF ATD Common Plan Representation ( which in turn may give input to the C2 Schema used in the JTF ATD.

A related information page is available here giving Austin Tate's input for the July 1996 Workshop.

A description of Tate's approach to Mixed Initiative Planning (mutually constraining the space of behaviour) and its relationship to the <I-N-OVA> constraint model of activity is available at here.

Here is briefing material on some background that we can draw on to improve the planning research input to the C2 schema. I am sure that much of this can be elaborated on, but I have tried to give the main sources of information that I would use as a starting point. Where possible, I point to on-line documents so we can begin the exchange of information promptly.


There are three types of plan representation or model we can draw on, and they are related.

At the May 1996 Annual Workshop of the DARPA/Rome Laboratory Planning Initiative (ARPI) in Edinburgh, the ARPI Plan Ontology Construction Group defined a framework for developing the various terminology bases for Air Campaign Planning work in ARPI. The model is as follows:

The following may look like a lot of input, but I hope you will see from the materials and from our work together that there is a great deal of agreement on some points, and some nice bases we can use for future protection.

A. Model-based Plan Ontology Input

A.1 Enterprise Ontology

A product of the Enterprise Project and fully published. English and KIF (in the Stanford KSL Ontolingua Library) versions available.

Probably the best overall "context" for activity schemas and models. It sets the organisational environment within which a good activity model can be defined. It deals with Organisation, Agents, external/internal relationships, etc.

See for pointers to the materials. Page 8 of the English language version provides a nice one page summary of everything in the ontology. The terms are non-military, but are meant to apply to ANY organisation with suitable interpretation of synonyms.

Jim Shoop of ISX and Austin Tate have already done a small informal comparison in December 1995 of parts of the C2 Schema as it existed in November 1995 against the Enterprise Ontology, and found that the Enterprise Ontology provided a useful frame of reference to look for omissions in the C2 schema, and to highlight structural anomalies, missing attributes, etc.

The activity and temporal models in the Enterprise Ontology were themselves based on KRSL, ARPI, new KRSL work in 1994, and Toronto TOVE work, as well as input from us. Hence they are very compatible with the general thread of the other inputs I list in section A.

A.2 Austin Tate input document for Enterprise Ontology, PIF and KRSL-Plans

The enterprise ontology includes a specific temporal model, but I view such parts as "add-ins" and the only commitment needs to be to a core shared model of some cross-relationship elements across the categories in the ontology. This point was difficult to get over during the Enterprise project, but is elaborated upon in my own inputs to the Enterprise ontology modelling and to the PIF and new KRSL-Plans efforts.

See or look at my Plan Ontology entry page for context and pointers to this and related material.

This document also includes a printed version of the May 1995 version of the KRSL-Plans new work which is listed next.

A.3 KRSL-Plans


Work following on from the 1991/2 efforts to define a language within ARPI suited to plan representation and interchange. This was KRSL 2.0.2. Some people used it but it was very complex and did not have a well underpinned plan/activity model.

A group of people in late 1994 and early 1995 set out to try to get back to the original idea behind KRSL, that of a genuine plan ontology. One outcome from a working group consisting of David Wilkins (SRI), Steve Smith (CMU), Bill Swartout (USC/ISI) and Austin Tate (AIAI) was the small KRSL-Plans model that is adopted by the Plan Ontology Group within ARPI Phase 3 now (though their main preoccupation is mostly with the "domain" lexicon/ontology which is covered under C below).

A.4 Process Interchange Format version 1.1


The core model is an elegant SMALL model largely consistent with the above mentioned references.

An important contribution of this work is the idea of "Partial Shared Views" used to extend the ontology as needed. This seems very important for any realistic uses. Other approaches such as Toronto's TOVE ontology and the inputs from Austin Tate (A.2 and A.6) also provide modularity and extension facilities by using a core small shared ontology.

A.5 International Workflow Management Coalition

See and especially the WfMC's Glossary of terms.

A draft of Working Group 1b's Process Definition Interchange standard has been provided for PIF/WfMC meta-model comparisons. Note that this has not otherwise yet been issued for comment by the WfMC.

This specification/interface is for process models and interchange. This is not yet published, so is not available to non-members. This is likely to have the broadest impact on commercial workflow and process representations in commercial products in the next few years. Working Group 1b doing the process model have been tracking PIF 1.1 and have stated that they believe the core models are similar enough that a merger is possible. The PIF working group meeting in Palo Alto in July 1996 will have the proposal of a merged standard as an item on their agenda.

A.6 <I-N-OVA> constraint model for activity

Issues - Nodes - Orderings/Variables/Auxiliary Constraints Model. See

<I-N-OVA> attempts to draw on work in disparate fields including practical and theoretical AI planning, SADT, IDEF, Issue based design, ISO standards work, etc. to propose a unifying theme for task/plan/activity representation. This is based on the idea of representing a set of constraints of various types on the behaviours possible in the modelled world.

The constraints may not all be amenable to CSP style reasoning and solving, but the unifying notion can help in communicating and manipulating the presentations of tasks, plans, activities, etc. <I-N-OVA> can support plug-in approaches to integrating model constraint management into systems (in a JTF ATD Reference Architecture compatible framework we use we have plug-in constraint managers of a range of different kinds). <I-N-OVA> also is intended to help with mixed user/system manipulation of plans, etc by providing a common reference point for work.

<I-N-OVA> goes further than most plan representation efforts, by explicitly recognising the need to represent the agenda of outstanding issues alog with a task, plan or activity. I think this is a vital element to transition us to modern workflow and work coordination approaches, and can be a help in defining mixed initiative systems.

A description of my approach to Mixed Initiative Planning (mutually constraining the space of behaviour) and its relationship to the <I-N-OVA> constraint model of activity is available at here.

<I-N-OVA> is being used to underpin a rationalisation of the Task Formalism (TF) task, activity and domain description language which has been used in the Edinburgh Nonlin and O-Plan planners for over 20 years - and which continues to be developed. The TF Manual is available at (377K).

B. Planning Capabilities Model Input

Section A deal with representing the "products" of planning. Also required is a way to represent the "process" of planning itself. There is much less work on this - though its now becoming an active topic.

B.1 Generic Task Representations from

B.1.1 CommonKADS Generic Task Library and recent work on planning tasks by Andre Valente (Amsterdam, now USC/ISI), John Kingston (AIAI)

B.1.2 EuroKnowledge's problem Solving Tasks

B.2 Recent work on ACP process models done within ARPI under the ISAT Project (ISX/AIAI) is available here

C. Domain Terminology or Lexicon (sometimes called Domain "Ontologies" by others)

This is the third and important part of the input. ARPI is creating a domain lexicon for ACP and already has a similar domain lexicon for transportation problems. Contacts are Bill Swartout ( and Lou Hoebel (