Introductions

[gref]

Hi, everybody!

I'm John Grefenstette, Head of the Machine Learning Group at the Naval
Research Lab in Washington.  Our Swarm team includes Bill Spears, Connie
Ramsey, Diana Gordon and Ken De Jong.  Our group has been active in
evolutionary computation for some time, and we're excited about
contributing to the Swarm effort.

We're looking at several areas, including:

1. Interactions between evolution and individual learning.  We've been
looking at evolution and learning here for quite some time.  We've
developed a system called SAMUEL that uses genetic algorithms on a
symbolic rule representation.  SAMUEL agents learn through a
reinforcement learning process, and also evolve over time using a
specialized genetic algorithm that contains both Darwinian and
Lamarckian components.  We've studied predator-prey tasks, navigation
tasks, and are currently developing some cooperative tasks for mobile
robots.  However, SAMUEL doesn't support hundreds of agents, so Swarm
might be very useful as a simulation testbed.  We've started our Swarm
orientation by implementing reinforcement learning in Swarm agents
(using a variant of the heatbugs model).

2. Models of emerging infections.  We're just starting a project with
the Walter Reed Army Institute of Research with the aim of building CAS
models of global emerging infections (e.g., AIDS), with the ultimate
goal of focusing control interventions at critical points of viral
evolution.  Swarm seems like a good vehicle for exploring hierarchical
models (e.g., cell/person/village/country/world) of emerging diseases.

3. Applications of CAS models to traffic control situations, including
air traffic control.

We'd like to hear about any other work by this group that relates
to any of these topics.

For more info on our work, please see http://www.aic.nrl.navy.mil.

Cheers,
John Grefenstette