FOUNDATION FOR INTELLIGENT PHYSICAL AGENTS

 

Document title:

ACL Maintenance Work Plan

Document number:

f-wp-00001

Document source:

FIPA Communication TC

Document status:

Approved

Date of this status:

2000/06/17

Change history:

2000/03/08

Initial draft

2000/04/05

Revisions made according to FAB comments

2000/06/17

FAB comments (see end)

 

                          Mark Greaves

                        <mark.t.greaves@boeing.com>

 

This work plan addresses FIPA’s basic responsibilities in the area of specifying an agent communication language, along with libraries of predefined communicative acts, interaction protocols, and content languages.

 

Problem Statement: Developers of multi-agent systems require communication frameworks in order to allow their agents to communicate with each other. Ad hoc systems are usually not sufficiently well designed or documented to be consistently extensible, implementable by others, or generally applicable to a wide set of agent problems.  The FIPA specifications for agent communication address these issues in ad hoc specifications.  The core of these specifications was largely completed in FIPA 97, but they have required continual maintenance and upkeep since then.

 

Objective: An important part of FIPA’s intellectual property is the definition and specification of a strong, formally grounded messaging system for inter-agent communication. This work plan covers the activity necessary to maintain and extend the specification documents for the logical form for such a system: the FIPA agent communication language.

 

Technology: This work plan will concentrate on developing, extending, and specifying the framework for an agent communication language based on the FIPA 99 specifications, and including predefined libraries of the following ACL elements:

·         Communicative acts,

·         Interaction protocols, and,

·         Content languages.

 

The results of this work plan will be embodied in a set of FIPA specification documents and library elements.  Specific core technologies that are relevant to these specifications include:

·         Logics (propositional, first-order, modal, dynamic, etc.),

·         Speech-act theory,

·         Protocol design and verification frameworks, and,

·         Messaging system semantics.

 

Specifications Generated: The goal of this work plan is to generate specifications in the following areas, at the Standard maturity level:

·         [FIPA00007] FIPA Content Languages Specification

·         [FIPA00025] FIPA Interaction Protocol Library Specification

·         [FIPA00037] FIPA Communicative Act Library Specification

·         [FIPA00061] FIPA ACL Message Parameter Library Specification

·         [FIPAacl] FIPA Agent Communication Language Specification

 

In addition, the three libraries contain specification documents which define content based on FIPA 99 and FIPA technology submissions, and will be managed in accordance with the procedures set out in the relevant administration document.

 

Plan for Work: The technical committee constituted by this work plan will take overall responsibility for producing the generated specifications. Because the specifications are tightly linked, various sub-committees of the overall agent communication technical committee will specialize in each area. Work to meet the plan goals will take place at FIPA meetings and throughout the year.

 

Specifically, we will:

·         Recast the FIPA 2000 ACL specification into a set of independently managed libraries for communicative acts, ACL message parameters, content languages, and interaction protocols.

·         Maintain these specifications and libraries by accepting and evaluating submissions from FIPA members and others.

·         Interface with the rest of the FIPA technical committees on issues of agent communication, and respond to problems and inconsistencies across the entire FIPA specification set.

 

Function as a collection point for feedback from implementations of the Experimental standards, and use this feedback to determine which of the Experimental standards to propose to the FIPA membership as Standard.

 

Milestones:It is anticipated that the work of generating the specifications in the FIPA 2000 format and bringing them to the Preliminary stage will be largely complete by 2000/04, and to the Experimental stage by 2000/07. Promotion of the specifications to the Standard level will depend on the progress of implementation work and the feedback received by this technical committee.

 

Future Work: Future work is anticipated in specifying non-BDI semantics, formally grounded accounts of interaction protocols, interaction protocol definition formalisms, limited-power and resource-aware logical frameworks, the linkage between large-scale agent policies and communication issues, ontology specification, and several other topics. This work will eventually be reflected in FIPA’s specifications for agent communication, and the technical committee constituted by this work plan will strive to be informed about developments in these and other areas. However, developing and/or standardizing these technologies will be addressed in a future work plan.

 

Dependencies: The specifications produced under this work plan depend on and interact with [FIPA00001] as well as the proposed work on policies. They also tightly interact with each other. These specifications are used by almost every other FIPA standard, and are a core part of FIPA’s standards set.

 

The individual library elements in this specification set may incorporate (by reference or inclusion) several external specifications for core technologies.  These include specific content languages and representation systems (e.g. KIF, XML), protocol specification formalisms (e.g. the OMG AUML spec), other agent communication frameworks (e.g. ARCOL, KAoS), and other agent-related technologies.

 

Support:

·         Luis Botelho, ADETTI

·         Jeff Bradshaw, Institute for Human and Machine Cognition/UWF

·         Mark Greaves, Boeing

·         Fabio Bellifemine, CSELT

·         Monique Calisti, EPFL

·         Tierry Bouron, France Telecom

·         Makoto Okada, Fujitsu Laboratories

·         Frank McCabe, Fujitsu Laboratories of America

·         Abe Mamdani, Imperial College

·         Jeremy Pitt, Imperial College

·         Jim Cunningham, Imperial College

·         Jim Odell, James Odell Associates

·         Nina Berry, Sandia National Laboratory

·         Bernhard Bauer, Siemens AG

·         Donald Steiner, Siemens TTB

·         Tim Finin, UMBC

·         Yannis Labrou, UMBC

 

FAB Comments:

The FAB have the following requirements to make in approving this work plan:

·         The work plan should only be concerned with the finalisation of the ACL specifications.

·         New work (that is, the introduction of new communicative acts, content languages and/or interaction protocols) should be proposed through a new work plan.

 

The FAB have the following recommendations for this work plan:

·         That the specifications are submitted to the FAB early enough so that they can be approved for Experimental status at the 2000/07 meeting. This would require one month prior submission to the membership and one month prior submission to the FAB; two months before the 2000/07 meeting.

·         That the library of components, in particular the content languages and the interaction protocols, also include recommendations to developers about how to choose between them, for instance, when should they use CCL instead of SL-1? Consider the differences between the descriptions of SL0, SL1, SL2 and SL as reported in the current standard [FIPA00008] and decide whether these help developers make a choice about which to use:

o        "Profile 0 of SL is the minimal subset of the SL content language. It allows the representation of actions, the determination of the result a term representing a computation, the completion of an action and simple binary propositions,"

o        "Profile 1 of SL extends the minimal representational form of SL0 by adding Boolean connectives to represent propositional expressions," and,

o        "Profile 2 of SL is a subset of the SL content language which still allows first order predicate and modal logic, but is restricted to ensure that it is decidable."

·         That this technical committee consider the possibility of merging SL-0 and SL-1 into a single language.

·         That this technical committee specify if, when and how a content expressed in content language X can be converted into content language Y. When it cannot, that the technical committee make clear in specifications that agents using content language X and content language Y can never interoperate.