ESF Scientific Programmes ALAMAS Outline Proposal in response to PESC’s Call for Outline Proposals 2002

OUTLINE PROPOSAL

ESF SCIENTIFIC PROGRAMME

Adaptive and Learning Agents and Multi-Agent Systems (ALAMAS)

Principal applicants:
Dimitar Kazakov*, / University of York, / UK
Daniel Kudenko, / University of York, / UK
Eduardo Alonso, / City University, / UK
*contact person
Summary:

Keywords: Agents, multi-agent systems, learning, adaptation, emerging behaviour

I. STATE OF THE ART, OBJECTIVES AND EXPECTED ACHIEVEMENTS

We propose the creation of a European Network on Adaptive and Learning Agents and Multi-Agent Systems (ALAMAS). The network will provide a suitable infrastructure for the development of this emerging multi-disciplinary area encompassing Computer Science, Software Engineering, Biology, Cognitive and Social Sciences, and support collaboration, specialisation and integration among European research centres. The creation of this network will help Europe keep its position as a leader in Agent Research, and, by preparing it for the new generation of agents able of adaptation and learning, boost its high technology industries, such as telecommunications, computer network management, multi-media and interactive entertainment, and increase quality of life through the development of personal assistants, …(more examples?), etc.

The notion of agents has been introduced to represent entities by their perceptions, actions, and behaviour relating the former to the latter. Conceptually, agents can be variously seen as descending from Systems Theory, Artificial Intelligence, Object Oriented Programming (OOP) or Social Sciences. Agents and Multi-Agent Systems (MAS) are powerful general-level abstractions used in the modelling of complex entities and systems. Recent research has put an emphasis on Intelligent Agents and MAS, which have become a highly active area of Artificial Intelligence (AI) research. Intelligent Agents have been developed and applied successfully in many domains, such as e-commerce, flexible networking, human-computer interaction, entertainment, process management and traffic control.

When designing agent systems, it is impossible to foresee all the potential situations an agent may encounter and specify its behaviour optimally in advance. Agents therefore have to learn from and adapt to their environment. This task is even more complex when nature is not the only source of uncertainty, and the agent is situated in an environment that contains other agents with potentially different capabilities and goals. Multi-Agent Learning, i.e., the ability of the agents to learn how to co-ordinate their actions, whether in the spirit of co-operation or competition, becomes crucial in such domains. The goal of this network is to increase awareness and interest in adaptive and learning agent research, encourage collaboration between experts from all relevant fields, and provide its community with a representative overview of current research in the area through a Web site, printed publications, and an annual scientific event. The network will serve as an inclusive forum for the discussion of ongoing or completed work in both theoretical and practical issues; it will also provide the backbone for a future integrated body of European research, which will help substantially to reduce the amount of redundant work and accumulate a critical mass of expertise and industrial contacts. The resulting synergy should speed up basic research, as well as help develop standard industrial scale approaches thus promoting the transfer of cutting edge know-how to industry.

The proposed Network will initially focus on (but is not necessarily limited to) the following topics: T1: Adaptive Mobile Agents, T2: From Single Agent to Multi-Agent Learning, T3: Learning of Co-ordination, T4: Learning and Communication, T5: Distributed Learning, T6: Evolutionary Agents and Emergent Multi-Agent Organisation and Behaviour, T7: Evolution and Individual Learning in MAS, T8: Game-Theoretical and Analytical Approaches to Adaptive MAS, T9: Logic-Based Agent Learning, T10: Learning in Reactive Agents, T11: Learning for Real-Time Applications, and, T12: Industrial and Large Scale Applications of Learning Agents.

II. EXPECTED BENEFIT FROM EUROPEAN COLLABORATION

It is expected that the initial effect of the creation of the Network will be a wider spread of existing know-how and the increase of the amount of joint research. This will gradually lead to the formation of medium and long-term teams of researchers with complementary expertise, and the specialisation of individual teams. In North America, research in ALAMAS is rife. In order to remain competitive, and keep its advantage in terms of existing expertise on Machine Learning (ML) and MAS, Europe needs to support research in the field, and ensure the dissemination of results through conferences, tutorials, books and online material. The multi-disciplinary nature of ALAMAS research requires inter-institutional exchanges of expertise to avoid duplication of effort and ensure effective dissemination of results. The proposed Network will support these tasks. The clusters of partners with interests in the same areas identified in the table in Annex 4 will be used to organise focused cluster meetings, which will make the discussions more effective and reduce travel cost while preserving seamless coverage of the whole research area.

Meetings with potential end users will provide suitable real-world test beds for the emerging technology, which will help compare results and bring research closer to the demands of industry. ALAMAS can assist the development of a new and sustainable Agent-Oriented Software Engineering Industry in Europe, contribute to the competitiveness of the thriving European computer games industry, and help develop a new generation of networking and telecommunications applications.

III. EUROPEAN CONTEXT

A number of European FP5 research initiatives – Esprit 7115 (MLNet), 29288 (MLNet2), BRA 6060 (ILP), LTR 20237 (ILP2) and [number] (AgentLink) – have helped establish Europe at the forefront of Agent and ML Research. The interest in combining these two and other related areas at a European level, first established by the AAMAS symposia ( has only recently been given a formal support by the Esprit XXX AgentLinkII NoE in the form of a designated Special Interest Group (SIG) (co-sponsored by MLNet2) on Agents that Learn, Adapt and Discover (ALAD). The ALAMAS proposal is built on the scientific and managerial experience drawn from AAMAS and ALADSIG. The proposed Network is needed, since despite its advances in the two most relevant areas of Agents and ML, Europe is lagging behind US in ALAMAS. It is also timely, as with the approaching end of the FP5 networks, a new European initiative will be needed to ensure that the momentum created by these is preserved and combined in order to achieve a maximum cross-fertilisation effect. The progress from an annual scientific forum (AAMAS) to a European SIG (ALAD) to a Network is natural and will help reduce the risk for the European financial support sought. The creation of a separate network is justified by the current level of interest (since its launch in Feb 2002, ALAD has become one of the largest AgentLink SIGs) as well as by the ALAMAS research agenda, which, unlike the rest of the Agent community, will need longer to develop industrial applications, and is mostly focused on middle to long term research. There is no doubt in the potential of the ALAMAS research, which, according to the currently drafted by AgentLinkII Roadmap, is expected to start playing a major role in Agent Technology around 2006-08. This period coincides with the second half of the proposed ALAMAS lifespan (2004-08) thus creating optimal conditions for establishing industrial contacts and developing large-scale applications in real domains.

IV. WORKPLAN, BUDGET ESTIMATE AND DURATION

The project’s proposed duration is five years: 2004-08. The total budget sought is €125,000/year with the following main headings:

  1. Annual workshop: €25,000/year
  2. Cluster meetings (6 per year, average number of partners: 5, expenses per participant: €800): €24,000/year
  3. Short visits: 25 * €800 = €20,000/year
  4. Research exchanges (up to a month): 15 * €2000 = €30,000/year
  5. Meetings with end users: €10,000/year
  6. Meetings of the Steering Committee: €5000/year
  7. Building and maintenance of Internet resources: €5000/year
  8. 5% administrative overhead: €6000/year

ANNEX 1: Applicants’ Contact Details and CVs

Dr Dimitar Kazakov
Affiliation and permanent address:
Artificial Intelligence Group, Department of Computer Science, University of York,
Heslington, York YO10 5DD
United Kingdom
tel. +44(0)1904 43 4775
fax: +44(0)1904 43 2767
E-mail: /
Contact until 15 Dec 2002:
Department of Intelligent Systems
Institute Josef Stefan
Jamova cesta 39
Ljubljana SI-1000
Slovenia
tel.: +386 1 477 3693
fax: +386 1 4251 038
Dr Daniel Kudenko
Affiliation and permanent address:
Artificial Intelligence Group, Department of Computer Science, University of York,
Heslington, York YO10 5DD
United Kingdom
tel. +44(0)1904 43 4776
fax: +44(0)1904 43 2767
E-mail: / Dr Eduardo Alonso
Affiliation and permanent address: Department of Computing, School of Informatics, City University
London EC1V OHB
United Kingdom tel: +44 20 7040 4049 fax: +44 20 7040 8587 E-mail:

ANNEX 2: Most Relevant Publications

ANNEX 3: Envisaged Steering Committee

Prof. Thomas Eiter / KBS Group, CS Dept., Vienna Univ. of Technology / Austria
Prof. Ann Nowé / Free University, Brussels / Belgium
Prof. Olga Stepankova / Czech Technical Univ., Prague / Czech Republic
Prof Luc De Raedt / University of Freiburg / Germany
Prof. Maria Teresa Pazienza / University Tor Vergata / Italy
Prof. Frances Brazier / Free University, Amsterdam / Netherlands
Prof. Eugenio Oliveira / University of Porto / Portugal
Dr Saso Dzeroski / Jozef Stefan Institute / Slovenia
Dr Enric Plaza / IIIA-CSIC / Spain
Dr Dimitar Kazakov (Chair) / University of York / United Kingdom

ANNEX 4: Candidate Participants

Name / Affiliation / Country
Prof Dr. Ann Nowé,
Prof Dr. Bernard Manderick
Tom Lenaerts
Anne Defaweux
Katja Verbeeck
Karl Tuyls
Sam Meas
Johan Parent
Jes Fink-Jensen
Joke Reumers
Piet van Remortel
Dr. ir. Kris Steenhaut
Lan Tran Ngoc / Vrije Universiteit Brussel
Faculty of Science
Department of Computer Science (DINF)
Computational Modeling Lab (COMO) / Belgium
Dr Peter Andras / Biological Computing and Bioinformatics Research Group, School of Comp. Sci., Univ. of Newcastle / UK
Dr Eduardo Alonso / Intelligent Agents Group, School of Computing, City University / UK
Dr Dimitar Kazakov / Artificial Intelligence Group, CS Dept., Univ. of York / UK
Dr Daniel Kudenko

ANNEX 4: Clusters of Participants and Topics of Interest

ANNEX 5: Previous ESF Applications

None.

ANNEX 6: Related Applications to the FP of EC or COST

An Expression of Interest (EoI) with the same title was submitted in response to the FP6 call in June 2002.

Contact person: Dr Dimitar Kazakov, , CS Dept., University of York, Heslington, York YO10 5DD, UK