FOSE-MAS

1
FOSE-MAS

• Session Wrap-UP

2
Major Points (1)

• Scott DeLoach
• We have yet to demonstrate benefits of MAS,
  address industrial needs
• Lack of key concepts, notations and models,
  industry strength methods
• Need to define a core meta-model for MAS
• Reduced costs, increased reliability
• Monique Calisti
• Need for technology transfer, interaction with
  industry
• Need solid, stable programming languages, tools,
  libraries, applications
• Focus on design, implementation and testing
• Good news SE gradually uses agents concepts
• Michael Winikoff
• Adaptiveness, complexity, goals more important
  than agents and autonomy
• Verification and validation are key issues
• In particular with dynamic structures, emergent
  behavior
• Document successful application
• Juan A. Rodriguez
• Autonomy, interaction, mobility will be key
  issues for future systems
• Need for agreement environment

3
Major Points (2)

• Paul Valckenaers
• Focus on domains where MAS can have an advantage
  over traditional SE
• Core business related applications, exhibit
  variability and heterogeneity
• Capture domain knowledge in reusable artifacts
  (cases, architectures, patterns, libraries)
• Reach out of the MAS community comfort zone
• Michael Georgeff
• Value proposition poor translation to mainstream
  SE
• Present agent concepts in traditional SE terms,
  show vlaue
• Bring key aspects/expertise of MAS SOA together
• Danny Weyns
• Do we ask the right questions for using MAS?
• Practical issues are a crucial concern
• Integrate MAS expertise in mainstream SE
  perspective
• Software architectures are important
• Aditya Ghose
• Agents are taken up in mainstream SE (e.g. SOA)
• Mediate MAS development in industry
• proactively connect with SOA
• But keep the agent label alive

4
Obstacles and Opportunities

• We are not aware of industrial needs and
  limitations
• State of the art of tools and methodologies far
  from realistic needs
• Incremental invasion into industry
• Lower development costs, ease of use
• Demonstrate benefits, cases of success
• Services as agents and vice versa
• Why use agents at all?

5
Summary (1)

• MAS-centered approach
• Need common MAS concepts
• Architectures, cases and patterns are essential
• Need accessible and powerful tools, languages and
  methodologies
• Account for non-agent elements, address NFR
• Practical aspects need attention testing,
  verification and validation
• Agents education in universities?

6
Summary (2)

• Relations between AOSE and other disciplines
• Combine key ideas from MAS and SE
• Work with SOA, Grid,
• Industry does not use what they do not need
• Demonstrate relevance to needs
• Cut costs, increase reliability, simplify
• No recognition by others, external communities
• We are not going to have MAS, but software
  systems (which may exhibit MAS properties)
• Meet requirements
• Down to earth

7
FOSE-MAS URL

• www.cs.kuleuven.be/danny/fose-mas.php
• Thanks!