Mehdi Dastani

1
PROMAS PROgramming Multi-Agent Systems

• Mehdi Dastani
• mehdi_at_cs.uu.nl
• Jorge J. Gomez-Sanz
• jjgomez_at_sip.ucm.es

2
Program

• 9 speakers in 7 sessions discussing various
  issues of MAS development
• social aspects, coordination, cooperation,
  decision
• reusability, reduce development costs, testing
• programming languages, tools, frameworks,
  libraries
• prototypes, experiments, case studies
• Session format
• 15 minutes talk 30-45 Discussion
• Joint meeting with AOSE

3
Tuesday 1 March (Afternoon session)

• Welcome and Round table introduction
  (14.15-14.30)
• Session 1 (14.30-15.30) AI and MAS- Ana García
  Serrano, Technical University of Madrid- Gabriel
  Marchalot, Thalès group, France
• Session 2 (15.30-16.15) Development/Maintenance
  Costs- Jorge Gomez Sanz, Universidad Complutense
  de Madrid
• Break (16.15-16.30)
• Session 3 (16.30-17.15) Imperative and
  Declarative Programming Languages for Multi-Agent
  Systems - Joao Leite, Universidade Nova de
  Lisboa
• Session 4 (17.15-18.15) Formal Methods for
  Multi-Agent Systems - Rafael Bordini, University
  of Durham, UK

4
Wednesday 2 March (Morning session)

• Session 5 (09.15-10.30) Multi-Agent Systems and
  Integrated Developing Environments - Gregory
  O'Hare, University College Dublin- Juan Manuel
  Serrano, Universidad Rey Juan Cerlos- Lars
  Braubach, Universität Hamburg
• Break (10.30-10.45)
• Session 6 (10.45-12.00) Implementation of
  Multi-Agent Issues - Birna van Riemsdijk,
  Utrecht Universiteit - Alessandro Ricci,
  Universita' di Bologna a Cesena
• Session 7 (12.00-12.45) Programming Pricinples
  for Multi-Agent Systems - Amal El
  Fallah-Seghrouchni, University of Paris 6
• Closing (12.45-13.00)

5
Session 1Artificial Intelligence and Multi-Agent
Systems

• How can AI and MAS areas be integrated?
• What is the influence of AI concerns in a MAS
  architecture?
• What is the experience in reusing AI libraries
  and software into MAS?
• Does AI really improve MAS?
• Can a MAS be built without AI?
• What can be learned from implementation and
  deployment of AI systems?
• Did AI need or result in special AI-oriented
  programming languages?
• Can AI-oriented programming languages be used for
  implementing multi-agent systems? In which way?

6
Session 2 Development and Maintenance Costs

• Is developing systems with agents less expensive
  in time and effort?
• How would you expect agent technology to decrease
  development costs?
• By increasing re-usability?
• By identifying concrete problems where agents,
  inherently, are better solution than other
  alternatives, say object oriented for instance?
• By producing more robust systems?

7
Session 3 Imperative and Declarative Programming
Languages for Multi-Agent Systems

• Imperative languages are chosen by most
  developers to produce their MAS.
• Can generic programming languages such as PROLOG,
  Oz, Haskell and Scheme be used to develop MAS
  systems? Is it really more helpful?
• What agent developments do you know which are
  implemented with declarative languages? Pure
  declarative implementations or hybrid?
• How can declarative languages be used together
  with imperative ones? How can declarative
  languages facilitate the MAS development?
• It seems that declarative languages can be
  applied to define the high-level behavior of an
  agent. Where else?

8
Session 4Formal Methods for Multi-Agent Systems

• There are significant amount of research in
  formal methods applied to agents, e.g. model
  checking approaches, modal logics, situation
  calculus.
• Formal methods are associated with high costs of
  developments since it requires time and high
  skilled workers.
• Is easier to apply formal methods to MAS than to
  other paradigms?
• what advantages would you remark when applying
  formal methods to a MAS development?
• Can agents be a better way of introducing formal
  methods deeper into industry?
• Are agents more easily accepted by industry due
  formal methods?

9
Session 5 Multi-Agent Systems andIntegrated
Developing Environments

• What are possible links between MAS integrated
  development environments (IDEs) and programming
  languages?
• Usually IDEs use a high level representation of
  the MAS that are translated later into executable
  code. What representations are more useful?
• what translation mechanism are considered?
• In the executable code, which programming
  language is more appropriate?

10
Session 6Implementation of Multi-Agent Issues

• Examples of MAS issues
• coordination, communication, mobility,
  environment, security, ontology, and services.
• Examples of single agent issues
• mental attitudes, agent roles, role dynamics,
  deliberation capabilities, adaptation
  capabilities, social capabilities, and reactive
  mechanisms.
• What are the techniques to implement these
  issues?
• Which languages provide programming constructs to
  implement them?
• How can these issues be implemented in existing
  programming languages such as Java or C?
• How can existing MAS architectures, which are
  defined in terms of higher level concepts, be
  implemented?

11
Session 7Programming Principles for Multi-Agent
Systems

• MAS development require training in different
  conventional disciplines, like object oriented
  programming, distributed systems, structured
  programming.
• Techniques used from other disciplines to MAS,
  like abstraction, inheritance, modularity,
  overloading, information hiding, error handling,
  generic programming, compositionality, and
  separation of concerns.
• What public implementations of MAS you know?
• What of the mentioned techniques you would
  highlight in those implementations?
• How would you value the importance or impact of
  those techniques in a MAS development?
• Do they influence more at Agent level or at MAS
  level?

12
Travel Grant

• Make sure your organization is an AgentLink III
  member
• AgentLink needs  actual expenditures (up to
  maximum)
• Original transaction receipts
• Send it before 1 April 2005 to AgentLink
•           http//www.agentlink.org/admin/expenses.
  html

13
Joint Session AOSE and ProMAS

• Is MOF a useful starting point for MAS
  programming languages or vice versa?
• Reverse engineering
• Which Semantics?
• Filling the gap between design Implementation
• Code generators
• Special Programming Languages and Platforms
• Testing Multi-Agent Systems
• Formal Methods
• Differences with Object-Oriented Testing
• Debugging Tools

14
Joint Session AOSE and ProMAS

• What are the following principles in MAS?
• abstraction,
• inheritance,
• modularity,
• overloading,
• information hiding,
• error handling,
• generic programming,
• compositionality, and
• separation of concerns

15
Joint Session AOSE and ProMAS

• Environment
• Platform, World, Services, resources
• Analysis (concepts) versus implementation
  (constructs)
• Coordination Mechanism
• Organization structure
• Analysis versus implementation
• Evaluation
• Relationship between Methodologies and
  Programming languages based on Implementations