MultiAgent System Development through SociallyBased Design

1
Socially-Based Design Meets Agent Capabilities
presenter Loris Penserini
(penserini_at_itc.it)
http//sra.itc.it/people/penserini/
A joint work with
Prof. Manuel Kolp University of Louvain
(BE) Prof. Maurizio Panti, Prof. Luca Spalazzi
Polytechnic University of Marche (IT)
2
Motivations
New application areas such as eBusiness, web
services, ubiquitous computing, and peer-to-peer
networks call for new requirements in software
development.
From the information system development
perspective, multi-agent systems (MASs) are a
powerful new software engineering paradigm for
designing and developing complex information
systems.
3
Motivations
Several agent oriented software development
methodologies have been proposed for analyzing,
designing, and developing complex systems, e.g.,
MaSE DeLoach et al., 2001, Gaia Wooldridge et
al., 2000, Tropos Castro et al., 2002, and
other proposals described in Iglesias et al.,
1998 and Caire et al., 2002.
Nevertheless, such methodologies do not strongly
concern with the important issue of describing
how the social aspects and the system
requirements can drive the definition and
characterization of the agent capabilities.
4
Goals
Taking advantage of an agent oriented
methodology, Tropos, we aim to characterize the
agent capabilities that better match the system
requirements.
To effectively deal with the architectural and
detailed design, we adopt a socially-based
approach.
Then, we give some examples in order to show how
and when the methodology high-level concepts
drive the formalization of intelligent agent
capabilities.
5
The Agenda

• - An overview of the Application Domain
• Organization based Architectural Design
• - Organizational roles as system actors
• Detailed Design by Social Patterns
• - The embassy pattern
• - The mediator pattern
• - mediator agent capabilities definition by means
  of agent abstractions
• - Conclusions and future directions

6
The Application Domain
We must develop a multi-agent system (MAS) to
support supply chain management scenarios.
Assumptions

• Let us assume that each enterprise is distributed
  on the network and has its own MAS.
• Every time an order occurs, the related
  enterprise can satisfy it in three principal
  ways a) using its internal stock, b) negotiating
  the quantity/lot required with the known
  partners, and c) trying to seek for new partners,
  i.e., a P2P interaction scenario.

7
The Application Domain
Principal architectural requirements

• The MAS has to manage the three kinds of
  collaboration according to its internal stock
  status
• The MAS has not to substitute the enterprise
  information system, but, it has to interface it
• The MAS has to allow the enterprise to
  collaborate even with new partners
• Each MAS have to be able to collaborate with
  other MAS at different levels of the supply chain
  network.

8
Organization-based Architectural Design
The Architectural Design, in Tropos, is conduct
by the use of organizational styles Kolp et al.,
2001 Kolp et al., 2003
Fig. 1 A MAS architecture based on the
organizational joint venture style
9
Organizational roles
Organizational roles are modeled by means of
system components in order to satisfy the system
requirements.
Facilitator. It is able to keep update the peer
actor with useful info on the partners addresses
and it exploits searching skills.
Information Source Manager. It coordinates
queries to information sources of the local peer
or/and of the other peers.
Supply Chain Manager. It coordinates
collaboration activities among distributed peers,
e.g., it exploits cooperation strategies to
overcome local failures.
10
Detailed Design
The main system components previously defined can
be modeled by means of the use of an agent based
approach.
According to the methodology, we have used social
patterns that are design patterns able to capture
social and intentional aspects that are recurrent
in multi-agent or cooperative systems Kolp et
al., 2001 Kolp et al., 2003 .
Moreover, social patterns allow designers for the
reuse of design experience and knowledge during
the whole system development process.
11
Social Patterns
Figure A) Peers Agent Platform as an embassy
social pattern
and B) Information Source Manager as a mediator
social pattern.
(A)
(B)
12
Agent Rationale
The relationships identified by social patterns
drive the agent capabilities modeling, i.e., the
agent-role characterization. For example, the
mediator agent calls for the following
capabilities Panti et al., 2001 Penserini et
al., 2002

• interpret the initiators request-message to
  correctly extrapolate the query
• reformulate the query according to its local
  knowledge
• cooperate with other agents to satisfy requests
• integrate the answers of a query, which are
  distributed and heterogeneous.

13
Agent Rationale capabilities analysis
In order to characterize the mediator
capabilities Panti et al., 2001, we exploit the
i Strategic Rationale model features Yu, 1995.
The main capabilities are Deal with
cooperation Deal with reformulation Interpret/Ex
ecute ACL performatives Answer integration
Fig. 3 Mediator capabilities analysis
14
Detailed Design needs agent abstractions
Agent capabilities allow the system to perform
decision-making activities at run-time.
Nevertheless, in order to support the
implementation phase of such capabilities, we
need to deal with the agent specification and the
knowledge representation techniques, at
design-time.

• For example,
• Agent interactions are often based on an
  instant-messaging communication model inspired to
  the speech act theory FIPA-ACL and KQML/KIF.
• Ontology definition plays a key role for the
  knowledge base characterization.

15
Capability Definition by Agent Specification
16
Agent Capability Formalization
17
Agent Capability Formalization
In our approach the Mediated Schema component is
also required to model the capability Deal with
cooperation.
Mediated Schema
18
ConclusionsFuture Work
This work presents an example of MAS development
emphasizing the need for organizational and
social abstractions to better capture the
architectural design aspects.
Moreover, we adopt the architectural and detailed
design results to drive the agent capability
modelling.
As future work, we are adopting DAML-S/OWL-S
specification to define the ontology in terms of
service annotations. Then, we are thinking to
relate the Tropos high-level concepts with the
OWL-S service descriptions.
19
Some References
Caire et al., 2002 G. Caire, F. Garijo, J.
Gomez, J. Pavon, E. Vargas, Agent-oriented
analysis using Message/UMLl, in Proc. of the 2nd
International Workshop on Agent-Oriented Software
Engineering, Vol. 2222 of LNCS, Springer
2002. Castro et al, 2002 J. Castro, M. Kolp and
J. Mylopoulos, Towards Requirements-Driven
Information Systems Engineering The Tropos
Project, in Information Systems, Vol. 27,
Elsevier, Amsterdam, The Netherlands, pp 365-389,
2002. DeLoach et al., 2001 S.A. DeLoach, and C.
Sparkman, Multiagent system engineering,
International Journal of Software Engineering and
Knowledge Engineering, Vol. 11, No. 3, pp.
231-258, 2001. Iglesias et al., 1998 C.
Iglesias, M. Garrijo and J. Gonzalez, A Survey
of Agent-Oriented Methodologies, Proceedings of
the 5th International Workshop on Intelligent
Agents Agent Theories, Architectures, and
Languages (ATAL-1998), pp. 317-330, Paris,
France, July 1998. Kolp et al., 2001 M. Kolp,
P. Giorgini and J. Mylopoulos. A Goal-Based
Organizational Perspective on Multi-Agents
Architectures, Proceedings of the Eighth
International Workshop on Agent Theories,
architectures, and languages (ATAL-2001),
Seattle, USA, August 1-3, 2001. Kolp et al.,
2003 M. Kolp, P. Giorgini and J. Mylopoulos,
Organizational Patterns for Early Requirements
Analysis, Proc. of the 15th Int. Conf. on
Advanced Information Systems Engineering
(CAiSE'03), Velden, Austria, June 2003.
20
Some References
Panti et al., 2001 M. Panti, L. Spalazzi, L.
Penserini, Cooperation Strategies for
Information Integration, in Proc. of the Sixth
Int. Conference on Cooperative Information
Systems (CoopIS), Springer Verlag, pp 123-134,
Trento, Italy, Sep 2001. Penserini et al., 2002
Penserini, L., Lin, L., Mylopoulos, J., Panti,
M., Spalazzi, L. Cooperation Strategies for
Agent-Based P2P Systems. Accepted in WIAS Web
Intelligence and Agent Systems An International
Journal, Publisher IOS Press, ISSN 1570-1263,
2002. Wooldridge et al., 2000 M. Wooldridge,
N.R. Jennings, and D. Kinny, The Gaia
Methodology for Agent-Oriented Analysis and
Design, in Journal of Autonomous Agents and
Multi-Agent Systems, 3(3)285-312, 2000. Yu,
1995 E. Yu, Modeling Strategic Relationships for
Process Reengineering, Ph.D. thesis, Department
of Computer Science, University of Toronto,
Canada, 1995.