Basics of MultiAgent Systems

1
Basics of Multi-Agent Systems
École dété FOR_at_C
Jean-Marc Frayret, Ph.D. and Luis Antonio
Santa-Eulalia, MSc.

• Mai 14th 2004

2
Content

• Global Objective
• Introduction / Context
• General concepts of agents
• Multi-Agent Systems
• Some Applications
• Final Remarks

3
Global Objective

• To present some basic high-level principles and
  concepts related to agents and Multi-Agent
  Systems (MAS), as well as to present some
  applications.

4
Content

• Global Objective
• Introduction / Context
• General concepts of agents
• Multi-Agent Systems
• Some Applications
• Final Remarks

5
Introduction / Context

• Engineering and computer science

6
Introduction / Context

• Agent technology to attend, at least, part of
  this old human desire

• Today they agent already show great value

7
Introduction / Context

• Agents beyond the automation

8
Content

• Global Objective
• Introduction / Context
• General concepts of agents
• Multi-Agent Systems
• Some Applications
• Final Remarks

9
General concepts of agents

• Definitions
• There is not a universally accepted and
  homogeneous
• An agent is a software system, located in an
  environment, and which acts in an autonomous and
  flexible way to achieve the goals for which it
  was conceived

10
General concepts of agents

• Agent paradigm VS. conventional systems

• Ex payroll system

11
General concepts of agents

• Agent State it can be its execution state and
  values of attributes
• Code necessary class to execute the agent
• Agent System platform that can create,
  interpret, execute, transfer and receive agents

Execution Unit Agent Computational Environment
System of Agents
12
General concepts of agents

• Other characteristics
• domain oriented reasoning
• sensing and acting
• goal oriented
• possibility to incorporate intelligence
• communication ability
• negotiating capacity
• collaborative
• self-starting
• temporal continuity
• character
• adaptive
• mobile

13
General concepts of agents

• Relations with the Environment

Wooldridge (1999)

• Ongoing and non-terminating action
• Repertoiry of actions

14
General concepts of agents

• Interaction of an agent with the environment and
  the interactions among agents

15
General concepts of agents

• Classification of environment properties

• Partial or total control of the environment

16
General concepts of agents

• Basic Internal Agents Organization

Transformation of agents data structure in
agents lifecycle
Dissimilar, identical, body-head agent
Similarity
Mutability
Agents may or may not retain a trace of changes
in their state based on their experience
Allows reuse of parts
Modularity
Memory
17
General concepts of agents

• Properties and classification of Agents

Adapted from Franklin, S. and Graesser, A.
(1996).
18
General concepts of agents

• Classification of Agents

Based on Franklin, S. and Graesser, A. (1996 )
19
General concepts of agents

• Classification of Agents

Based on Franklin, S. and Graesser, A. (1996 )
20
General concepts of agents

• Intelligent Agents
• Polemic theme
• Capacity to react rationally to a stimuli from
  the environment
• In a unpredictable or open environments
• Where there is a significant possibility that
  actions can fail
• Flexibility and adaptability
• Ability to represent and manipulate knowledge

21
General concepts of agents

• Mobile Agents

Mobile
22
General concepts of agents

• Client-Server vs. Mobile Agents

23
Content

• Global Objective
• Introduction / Context
• General concepts of agents
• Multi-Agent Systems
• Some Applications
• Final Remarks

24
Content

• Global Objective
• Introduction / Context
• General concepts of agents
• Multi-Agent Systems
• Definitions
• Communication
• Development
• Limitations and challenges
• Some Applications
• Final Remarks

25
Multi-Agent Systems

• Broad the concept of individual agent
• Definition
• a set of agents that work together and interact
  in order to accomplish some tasks
• they use their competences and knowledge to
  strengthen the capacity of solving problems

26
Multi-Agent Systems

• Characteristics
• Each agent has limited capacities and information
  of problems resolution
• Each one has a partial point of view
• The MASs have no global control
• All data are decentralized
• All calculations are asynchronous

27
Multi-Agent Systems

• Advantages

Transformation of agents data structure in
agents lifecycle
High speed
Agents may or may not retain a trace of changes
in their state based on their experience
Allows reuse of parts
Modularity
Reliability
28
Multi-Agent Systems

• Some Important Mechanisms
• Interaction
• Cooperation
• Coordination
• Negotiation
• Planning
• Communication

29
Content

• Global Objective
• Introduction / Context
• General concepts of agents
• Multi-Agent Systems
• Definitions
• Communication
• Development
• Limitations and challenges
• Some Applications
• Final Remarks

30
Multi-Agent Systems

• Communication Models
• Blackboard Schema

31
Multi-Agent Systems

• Communication Models
• Direct Exchange of Messages

Adapted from Lucena (2003 )
32
Agent Communication Languages

• Type of messages to support the communication
  process
• Two main ACLs
• KQML (Knowledge Query and Manipulation Language)
• ACL (Agent Communication Language) from FIPA.

33
Knowledge Query and Manipulation Language (KQML)

• Main objective to allow the knowledge sharing
  among applications
• High level
• Allows exchange of messages independent of

KQML
34
KQML Performatives

• Based on a set of performatives
• represent the intention of the agents when
  sending some message

(ask-if (lt (size chip1) (size chip2)))
(reply true)
35
KQML Basic Example
(tell sender A receiver B content price
(ISBN1234567890, 24.59) language Prolog
ontology ecommerce in-reply-to message IDxy123)
36
Agent Communication Language (ACL) from FIPA

• FIPA (Foundation of Intelligent Physical Agents)
• Benefited from many technological evolution of
  KQML
• Incorporates a lot of instruments to treat the
  semantic requirements

• The syntax is similar to KQML, but the
  performatives can be different

37
ACL/FIPA Basic Example
(inform sender A receiver B content price
(ISBN1234567890, 24.59) language Prolog
ontology ecommerce in-reply-to xy123A
conversation ID xy123A)
38
Knowledge Representation Languages

• Inference machines
• able to process knowledge stored in the knowledge
  BD and interpret it

39
Ontology Definition

• It provides a machine-processable semantics of
  information sources
• Easing the communication between agents

• Definition (Gruber, 1995)
• A formal explicit specification of a shared
  conceptualization
• Conceptualization an abstract model of some
  phenomenon
• Explicit the type of concepts used and the
  constraints on their use are explicitly defined
• Formal should be machine-readable
• Shared consensual knowledge

40
Ontology

• It provides a machine-processable semantics of
  information sources
• Easing the communication between agents
• Facilitate the construction of a domain model
• Proving
• A vocabulary of terms
• Specification of its meanings
• Relations
• Usually organized in taxonomies

41
Ontology Example

• 4 levels Classification Hierarchy
• Hierarchy Segment / Family / Class / Commodity
• Representation NN.NN.NN.NN
• Example

42
Content

• Global Objective
• Introduction / Context
• General concepts of agents
• Multi-Agent Systems
• Definitions
• Communication
• Development
• Limitations and challenges
• Some Applications
• Final Remarks

43
MAS Development

• MAS are considered complex systems that deserve
  great effort to develop it

Macro Issues
44
Micro Issues

• Best-known agent architecture is the Procedural
  Reasoning System

(Woodridge, 1998).
45
Macro Issues

• Objective How one designs an agent society that
  can (co)operate effectively

• Societies, not individuals

• Contract Net
• The best-known framework for DPS

46
Contract Net
(Woodridge, 1998)
47
DPS MAS

• MAS
• Societies of autonomous agents
• Share a common goal not always
• They can conflict their interests
• Questions coherence, coordination,
  communication, cooperation, and negotiation

Distributed problem solving (DPS) methods
48
Agents, Objects and Expert Systems

• Agents and OO approach
• Autonomy, Flexibility, Multi-thread
• Sociability, pro-activity and reactivity
• Is possible to implement agents using OO tech
• New paradigm agent-oriented approach

• Agents and expert systems (ES)
• ESs do not sensor (not acting directly)
• ESs do not act on any environment
• Are not necessarily capable of co-operating

49
Content

• Global Objective
• Introduction / Context
• General concepts of agents
• Multi-Agent Systems
• Definitions
• Communication
• Development
• Limitations and challenges
• Some Applications
• Final Remarks

50
Some Limitations

• A lot of applications that use agents can be
  developed using other techniques

51
Some Limitations

• More?...
• Agents do not have complete global Knowledge
  about its environment
• Globally sub-optimal decisions are common
• It can take time until users gain confidence in
  the agents.

52
Main Challenges
53
Main Challenges
54
Content

• Global Objective
• Introduction / Context
• General concepts of agents
• Multi-Agent Systems
• Some Applications
• Final Remarks

55
Some Applications

• Distributed systems
• Human-computer interface
• Distributed and cooperatives databases and
  knowledge bases
• Systems for the comprehension of the natural
  language
• Communication protocols and telecommunication
  networks
• Agents oriented programming and software
  engineering
• Cognitive robotics and co-operation between robots

56
Some Applications

• MAS inspires studies related to diverse
  disciplines, in particular
• sociology, social psychology, cognitive sciences
  and others

57
Agents in Our Daily Life
58
Industry Applications

• Process control
• Manufacturing planning, scheduling and control
• e-Procurement
• cade study

59
Consortium
Brazil
Portugal
Poland
Spain
Greece
Supported by
England
Belgium
EU
Brazil
60
The Deepsia Project General Vision

• DEEPSIA Acronym
• Dynamic on-linE IntErnet Purchasing System based
  on Intelligent Agents IST Programme
• Objectives
• Aims at addressing the purchasing business
  process within SMEs with an e-Commerce
  application
• Helping to perform usual day-to-day purchasing
  tasks using the potential of the WWW

61
The Deepsia Project Basis (1/2)

• Multi Intelligent Agent System
• Autonomously generation of an electronic
  catalogue of products

62
The Deepsia Project Basis (1/2)

• Normalization of product data from multiple
  vendors
• So it can be easily compared
• Is expected to achieve less costly and more time
  effective purchases
• SMEs seen as buyers

63
The Multi Agent System MAS

• 2 basic mechanisms to collect information

64
DEEPSIA Architecture
http//www. .com
Multi
-
Agent System
Multi
-
Agent System
65
Content

• Global Objective
• Introduction / Context
• General concepts of agents
• Multi-Agent Systems
• Some Applications
• Final Remarks

66
Final Remarks

• We have introduced some basic concepts
• Why agents go further than the conventional
  software
• Some difficulties to develop a MAS
• No general methodology
• No commercial tool
• There is an enormous variety of texts available
• A lot of challenges and opened questions

67
Thank you !
Contact Jean-Marc.Frayret_at_forac.ulaval.ca Luis.An
tonio.Santa.Eulalia_at_centor.ulaval.ca http//www.fo
rac.ulaval.ca
68
Main References 1/3

• Deepsia Consortium. Technical Deepsia Annex 1
  Description of Work. July/2000. 87p. Report. IST
  PROJECT-1999-20483.
• Bentahar, J., Moulin, B. Chaib-draa, B..
  "Towards a Formal Framework for Conversational
  Agents". In Agent Communication Languages and
  Conversation Policies AAMAS 2003 Workshop,
  Melbourne, Australia, 14 july 2003.
• ECCMA (Electronic Commerce Code Management
  Association) home page. Available at URL
  http//www.eccma.org. Last visit at 18/04/03.
• FENSEL, D. 2001 Ontologies A Silver Bullet for
  Knowledge Management and Electronic Management.
  Berlin Springer.
• FIPA (Foundation for Intelligent Physical Agents)
  home page. Available at URL http//www.fipa.org
  /. Last visit at 16/04/2004.
• Franklin, S. and Graesser, A. Is it an Agent,
  or just a Program? A Taxonomy for Autonomous
  Agents. Proceedings of the Third International
  Workshop on Agent Theories, Architectures, and
  Languages, Springer-Verlag, 1996.
• GRUNINGER, M. LEE, J. 2002. Ontology
  applications and design. Communications of the
  ACM. v.45, n.2, February.

69
Main References 2/3

• Jarras, I and Chaib-draa, B. Aperçu sur les
  systèmes multiagents. Séries Scientifique.
  CIRANO Centre Universitaire de Recherche en
  Analyse des Organisations. Montréal, juillet
  2002.
• Jennings, N. R. Wooldridge, M. Applications
  of Agent Technology. In N. R. Jennings and M.
  Wooldridge, editors, Agent Technology
  Foundations, Applications, and Markets.
  Springer-Verlag, March 1998.
• Jennings, N.R. Sycara, K. Woodridge, M. A
  Roadmap of Agent Research and Development.
  Autonomous Agents and Multi-AgentSystems, 1
  ,738, 1998.
• Lucena, P. Semantic agent uma plataforma para
  desenvolvimento de agentes inteligentes.
  Masters Thesis, ICMC-USP, Brazil, 2003.
• Milagres, F. G. Moreira, E. S. Pimentão, J. P.
  Sousa, P. A. C. Garção, A. S. Dealing with
  Security within DEEPSIA Project In. WSEAS int.
  conf. on information security, hardware/software
  codesign, e-commerce and computer networks, 2002,
  Rio de Janeiro Proceedings of the WSEAS Int.
  Conf. on Information Security, Hardware/Software
  Codesign, E-Commerce and Computer Networks, WSEAS
  Press, 2002, p. 2431-2439

70
Main References 3/3

• Parunak H.V.D. Practical and Industrial
  Applications of Agent-Based Systems.
  Environmental Research Institute of Michigan
  (ERIM), 1998.
• Santa Eulalia, L.A. Moreira, E.S Carvalho,
  A.P.L.F. Rozenfeld, H. (2002) Using ontologies
  for intelligent agent trainning and information
  retrieval in a e-commerce application case study.
  In ECCPPM 4th European Conference On
  Product And Process Modeling. Proceedings.
  Portoroz, Slovenia, Setember 2002, A. A. Balkema
  Publishers, Lisse, The Netherlands. pp. 277284.
• Smith, R. The contract net protocol. IEEE
  Trans. on Computers, 1980. C-29(12), 1104- 1113.
• W3C Consortium home page. Available at URL
  http//www.w3.org/. Last visit at 01/03/04.
• Wooldridge, M. Agent-based computing. In
  Baltzer Journals. September 29, 1997.
• Wooldridge, M. Intelligent Agents. In G.
  Weiss, editor Multiagent Systems, The MIT Press,
  April 1999.