Dear AI/Games Researcher,

Funded by a Technology Strategy Board feasibility study grant,
via a partnership of Introversion Software and Imperial College,
we have developed an Application Programming Interface (API) for
Introversion's DEFCON game. DEFCON is a very popular nuclear war
strategy simulation game, details of which can be found here:
www.introversion.co.uk/defcon

DEFCON is a full-featured highly graphical commercial game, but
because it is also a strategy game, it is amenable to
experimentation with AI techniques. In particular, DEFCON comes
with a 1-player mode where you can play against an automated
opponent (called an AI-bot). For his masters project, Robin
Baumgarten - who is the lead software engineer building the API -
wrote an alternative AI-bot which was able to beat the original
AI-bot in three out of every four games. Robin's approach used a
combination of reasoning techniques including re-inforcement
learning, simulated annealing and case-based reasoning. This was
a good start, but there is still much room for improvement, and
we hope that DEFCON will be as big a driving force for the
development of AI techniques as chess or backgammon was.

And this is where the API comes in. The API will allow developers
to control every aspect of an AI-bot's behaviour as it plays the
game against a human opponent (or another AI-bot). In particular,
it will allow intelligent placing of resources such as missile
silos and air-bases, and it will allow fine-grained control over
bombing missions and fleet movements. The API will allow
programmers to use Java, C++ or Python to interact with the game,
and we will supply an example AI-bot which can be the basis of a
new AI-bot, in addition to tutorial material on using the
API. Robin will be available for some technical assistance in
getting the API working, and will of course fix any bugs if they
arise.

The API will be available from early November from this site:

http://www.introversion.co.uk/defcon/bots/index.html

On the site will also be tutorial information, and access to
Robin's MSc.  dissertation and a couple of papers which resulted
from his work.

We would like to encourage you to set student projects which
involve using the API to develop AI-bots for DEFCON. Here is a
great incentive for students to do well: we plan to hold a DEFCON
competition at the 2009 IEEE Symposium on Computational
Intelligence and Games. Can your student write the world's most
advanced AI-bot to play DEFCON? There are many AI techniques
which could be used to control the overall strategy or individual
aspects such as fleet movement.  These include behaviour trees,
machine learning (in particular, we think that ANNs could be put
to good use), constraint solving and evolutionary techniques. To
highlight this, I will be supervising a student project this year
which involves evolving behaviour trees.

If you have any enquiries about the API and how you could use it,
please contact us. In particular:

For technical enquiries about the API, please contact Robin
Baumgarten: robin.baumgarten06@doc.ic.ac.uk

For enquiries about DEFCON and Introversion Software, please
contact Mark Morris: mark@introversion.co.uk

For enquiries about using the API for student projects, please
contact me (Simon Colton): sgc@doc.ic.ac.uk

We hope that the API will be of use to you for interesting
student projects.  Please pass this email onto researchers you
think might be interested.

We are putting together a mailing list for occasional updates
about the API, including details of the first release date - due
around 7th November. There is a link to this available from the
web page given above - please join up.

All the best,

Simon Colton
---
Senior Lecturer, Department of Computing, Imperial College, London
http://www.doc.ic.ac.uk/~sgc