From animals tot animats 7
by Hallam,
Bridget, Dario Loreano, John Hallam, Gillian Hayes and Jean-Arcady Meyer
(eds.)
MIT Press, Cambridge MA, 2002,
432 pp., illus. b/w, paper, $70.00
ISBN: 0-262-58217-1
Reviewed by Stefaan Van Ryssen
Hogeschool Gent
Jan Delvinlaan 115, 9000 Gent, Belgium
stefaan.vanryssen@pandora.be
These are the Proceedings of the Seventh International Conference on
Simulation of Adaptive Behavior held in Edinburgh in August 2002.
The editors of this book, and probably most of the contributors to the
conference and the book as well, believe that the best road to understanding
adaptive behavior in uncertain environments is through simulation. In
devising and evolving rules for the behavior of not so simple computational
systems, we gain insight in the underlying mechanisms of those adapting
and evolving systems we call "life". And that is what it is all about.
Ecologists, engineers, ethologists and economists believe they can profit
form those insights, as well as scientists involved in artificial intelligence,
robotics and psychology. To get to the animals, and ultimately to ourselves,
we turn to animats. In this sense the title could be read backwards
as well: from animats to animals.
However, most of the papers in this book deal with the first swing of
the pendulum. Trying to simulate really existing behavior or some subset
of all possible behaviors - mostly the more easily formalised subset
- of an organism is an enormous challenge and we have a long way to
go to fully understand the complexities of the interplay between an
organism, its way to perceive the world, the border conditions of its
own functionality and the resulting emergent behavior. The other swing
will be more daunting still: identifying and understanding the behavior
of animals through the multiple analogies with our own creations.
The book is organised in seven themes and contains some seventy papers,
carefully selected in a two-tier review process from the 139 presented
at the Conference.
'The Animat Approach to Adaptive Behavior' has two must-read papers.
Mark Humphrys and Ciaràn O'Leary briefly describe the WorldWideMind
(WWM) as an architecture for a distributed interdisciplinary research
intelligence. They propose the outlines for a unified mind, constructed
from many remote components written by several oauthors. 'The aim is
to address the scaling up of animat research, or how to construct minds
more complex than could be written one author (or one research group).'
In a second part of the paper, the first implementation of this WWM
is described and a way to write a WWM server is explained.
The exciting thing about this proposition is that, when the WWM is succesful,
it will be in itself a low level simulation of a more complex adaptive
system.
Anil K. Seth defends the idea that 'behavioral and/or mechanistic complexity
can be understood in terms of mediating well-adapted responses to environmental
variability.' The paper profoundly discusses the framework within which
ther relations between behavior, mechanism and environment can be usefully
articulated. Several examples of agent-based modelling techniques complement
this framework.
The other themes of the book are: Perception and motor control; Action
selection and behavioral sequences; Internal world models and processes;
Self-organization and learning; Evolution; and Collective and social
behavior.
It is obvious that this is not a book for someone who wants to get a
consistent and easy-to-read introduction in the problems and possibilities
of the simulation of adaptive behavior. Instead of the full picture,
which I believe no one at present would be able to paint, the reader
gets glimpses from very different angles. Browsing through the papers
under the different themes is very rewarding though. With some effort,
you will get an idea of what the elephant looks like and what the beast
has in store for the years to come.