** CALL FOR PAPERS **

H-AAMAS: Hierarchical Autonomous Agents and Multi-Agent Systems
Half-day workshop at the AAMAS-06 conference
May 9 (afternoon), 2006
http://www.science.uva.nl/~bram/HAAMAS/

Submission & notification:
--------------------------
Submissions must be sent to bram (at) science.uva.nl no later than
*February 1, 2006*.
Notification of acceptance or rejection will be sent no later than
*February 19, 2006*.

Workshop description:
---------------------
In a variety of fields related to autonomous agents and multi-agent
systems, hierarchical approaches are beginning to emerge as one of the
premier ways of dealing with the scale and complexity of interesting,
real-world problems. These fields include reinforcement learning,
evolutionary algorithms, multi-agent learning, mapping and planning in
robotics, Markov processes, and networked sensor and information systems.

The general strategy in all these methods is "to divide and conquer": a
large, complex problem is decomposed (possibly recursively) into
smaller, simpler subproblems. Hierarchical methods generally represent
and solve tasks at multiple spatial and/or temporal resolutions, and
higher levels or layers are in some sense abstractions of the details of
lower levels. However, precise, formal relationships between
hierarchical methods in different fields are virtually unknown, while
presumably hierarchical methods in one field may profit greatly from
advances made on hierarchical methods in another field.

The purpose of this workshop is to bring together researchers from these
different fields to discuss the similarities between the various
hierarchical methods, inspire cross-fertilization, prevent superfluous
reinventions of the same "hierarchical wheels", and identify important
questions for further research on hierarchical methods.

Call for papers:
----------------
We invite papers on hierarchical methods, in particular related to the
following, but not limited to the following issues:

*Which mathematical frameworks can provide the solid theoretical
underpinnings of hierarchical methods?
*How can hierarchical methods developed for single agent systems be
extended to multi-agent systems?
*There are some results on the complexity of hierarchical problem
solving in evolutionary algorithms. Can we generalize those results to
other fields in which hierarchy is important?
*What is the relationship between hierarchical methods and "flat",
"monolithic" methods which do not exploit hierarchical structure, both
in terms of the type of solutions and in terms of the savings that can
be obtained with hierarchical methods?
*There exist mathematically sound methods for learning policies given a
hierarchical structure, for example in reinforcement learning. Can we
provide similarly sound methods for learning the hierarchical structure
itself?
*Hierarchical solutions sometimes perform suboptimally. Can these losses
as a result of hierarchy be formalized, and can solutions be guaranteed
that are optimal given the hierarchical structure?
*What are the most important questions future research on hierarchical
methods should address?
*Applications of hierarchical methods on complex/real-world problems.

Workshop organizers:
--------------------
Dr. Bram Bakker
Intelligent Autonomous Systems
Informatics Institute
University of Amsterdam, The Netherlands
bram (at) science.uva.nl

Dr. Leon Kester
Integrated Systems
TNO Defense, Safety and Security, The Netherlands
leon.kester (at) tno.nl

Dr. Nikos Vlassis
Intelligent Autonomous Systems
Informatics Institute
University of Amsterdam, The Netherlands
vlassis (at) science.uva.nl

Dr. Edwin de Jong
Department of Information and Computing Sciences
University of Utrecht, The Netherlands
dejong (at) cs.uu.nl