MIME-Version: 1.0 Content-Type: multipart/related; boundary="----=_NextPart_01C67F1D.7D2CB960" Questo documento è una pagina Web in file unico, nota anche come archivio Web. La visualizzazione di questo messaggio indica che il browser o l'editor in uso non supporta gli archivi Web. Scaricare un browser che supporti gli archivi Web, come Microsoft Internet Explorer. ------=_NextPart_01C67F1D.7D2CB960 Content-Location: file:///C:/0C4BB24E/AgentSpecification.htm Content-Transfer-Encoding: quoted-printable Content-Type: text/html; charset="us-ascii" 1

&n= bsp;

Version: January 10, 2006

Document Authors: Massimo Cossentino a= nd Valeria Seidita=

Index

1 Introductio= n. 1=

2 Fragment Description. 2=

2.1 = Portion of = process. 3=

2.1.1 &= nbsp; Process rol= es involved. 5=

3 Deliverable= s. 5=

3.1 = Preconditio= ns. 5=

3.2 = Guideline(s= ) 6=

3.3 = Aspects of = Fragment 6=

4 Relationshi= ps with the MAS Meta-model 6=

5 Dependency relationships with other fragments. 7=

6<= /span> Input/Output<= /span> 7=

7<= /span> Glossary. 8=

<= span style=3D'mso-bookmark:_Toc102126676'>1&nb= sp; Introduction

ADELFE (Atelier de DÉveloppement de Logiciels à Fonctionnalité= Emergente) is a design process fo= r the adaptive multi-agent software systems developing; it is based on the AMAS theory.

Adelfe design process is shown in (= see Figure 1) in terms of its composing phases (requirements, analysis and design) and the work products produced:

- = Requirements phase produces the Environment Model;

- = Analysis phase produces the Identification of the Agent Model;

- = Design phase produces the Agent and the NCS Models.

The requirements phase prov= ides the Environment Model that, in the AMAS theory, is used as base for the pro= cess of adaptation. This process begins with the interactions between the system= and the environment, which includes <= /span>the following activities: actors determination, context definition and environm= ent characterization.

In the analysis phase, the previously defined entities are analyzed in order to specify which will be agents. An agent is an entity having the capability to evolve during an unexpected situation, showing new behaviours= and skills.

The design phase aims to defin= e the agents’ architecture describing their behaviours.

Figure 1. The complete Adelfe process

In this document we will define the Agents Specification fragment extracted fr= om Design phase.

<= span style=3D'mso-bookmark:_Toc102126676'>2&nb= sp; Fragment Description

The process to be performed in this phase is described in the following Figure 2. It is composed of three sub-phases level work definitions (Architecture Definition, Agents Specification and Architecture Refinement) and several related work products (mainly UML models and text documents).

Figure = 2. The Design phase described= in terms of work definitions and= work products

Let us consider the Agents Specification work definition (the blue oval), with its work products. This fragment, for each agent identified in a previous phase, aims at defining i= ts behaviour: the skill, the aptitudes, = an interaction language, a world representation, the Non Cooperative Situation= s, finally the identified behaviours are tested .

2.1 Portion of process

The process that is to be pe= rformed in order to obtain the result is represented in Figure 3 as a SPEM diagram:

Figure = 3. Agents Specification – Procedural aspect

Activity	Activity Description	Roles involved
Study Interact= ion Languages	The agents interaction lan= guages are defined	Agent Designer= (perform)
Design Agents<= o:p>	The agents behaviors are defined	Agent Designer= (perform)
Fast Prototypi= ng	The agents behaviors are tested	Agent Designer= (perform)

<= o:p>

2.1.1 Process roles involved

2.1.1.1 Agent Designer

He is responsible of:

To study Interaction Languages during the Agent Specification work definition. The agents interaction languages are defined.
To design Agents during the Agent Specification work definition. The agents behaviors are defined.
Fast Prototyping during the Agent Specification work definition. The agents behaviors are t= ested.

<= o:p>

3 = Deliverables

The protocols that specify t= he interaction languages are represented through AUML diagrams and they are described in the Protocol Diagrams. describes the relationship of the fragment with respect to the MAS model:

= ;

Figure 4. MAS Metamodel concept

<= o:p>

5 = Dependency relationships with other fragments

This fragment depend on two fragments:

Agent Identification, because the agents are defined in this fragment;<= /o:p>
Architecture Definition, because the agent specification uses the Detailed Architec= ture (initial) document.

6&nb= sp; Input/Output

Input, output and element to be designed in the fragment are detaile= d in the following table:

INPUT	TO BE DESIGNED	OUTPUT
Detailed Architecture (ini= tial)	Interaction Languages=	Interaction Languages (ini= tial), Protocol Diagram
Detailed Architecture (ini= tial), Interaction Languages (initial)	Agents Behaviors	Interaction Languages (dra= ft), Detailed Architecture (draft)
Interaction Languages (dra= ft), Detailed Architecture (draft)	Tested Agents Behaviors	Interaction Languages (fin= al), Detailed Architecture (including agent model)

7 Glossary<= /a>

Agent - Physical or virtual entity that it possesses the following abilities: to ac= t in an environment, to communicate with other agents, to pursue an objective, to reproduce itself, to possess resources, to perceive his environment, to have one limited (or absent) environment representation.

Adaptive Multi-Agent System AMAS - The behavior= is chosen while running to complete the task. It aims to get the better result= .

Aptitude - Agent characteristic to reason about its knowledge and beliefs.<= /span>

Characteristics - Agent intrinsic or physical property.

Cooperation - The cooperation is a social attitude of the agent that allows it to detect = and to resolveNon Cooperative Situation (NCS), as incomprehension, ambiguity, e= tc.

Cooperation <= /span>failure – NCS detection, it can occur when the coop= eration protocol isn't respected or wrong interactions are occurred between the sys= tem and its environment.

Cooperative agent - A cooperative agent is an agent with a social and cooperative attit= ude. Its lifecycle is classical perceive-decide-action.

Entity – An entity is a set of roles that the users play when they interact with use= cases.

The entities can be of two t= ypes:

Active entities that act autonomously;
Passive entities that can only exchange data with the system, therefore, they = are considered system resources.

Environment - The environment is everything that surrounds the agent in which this acts.<= o:p>

Goal - A goal is the objective that the system must to achieve and to maintain.=

Interaction - The languages of interaction allows to an agent to communicate with other agents or directly with its environment. The interaction can be classified = in: perception and action.

Non Cooperative Situation - NCS are unusual situations in which the agent must face an unpredictable environment. Th= ere are different kinds of NCS: incomprehension, ambiguity, incompetence, unproductiveness, concurrency, conflict, uselessness.

Perception - Through the perception is possible to receive some information on the physi= cal or social environment, therefore the designer must to endow the agents of perceptive ability.

Representation - A representation is used from the agent to determine its behaviour.= The representation that evolve are represented through the multi-agent system.<= o:p>

Skill - The skill of an agent is referred to the knowledge that it has on the domain, allowing him to understand what are its functions.

<= o:p>