Re: simulation and counterterrorism

[Marcus G. Daniels]

>>>>> "RR" == Reid, Reyhan <address@hidden> writes:

RR> I just stumbled accross the SWARM website in my attempts to locate
RR> analysis tools that model terrorism, assymetric warfare, and the like. 

I think agent-based simulation would be a good approach to the problem.

It's a question of implementing plausible behaviors (e.g. terrorist
targets/strikes and military/inter-government interventions) to a
sufficiently realistic degree, having a realistic landscape (e.g.
locations to strike with particular vulnerabilities that both
governments and terrorists observe), and having some outcome metrics
to optimize over.

I'm imagining a kind of wargaming platform for measuring a
multidimensional probability distribution of outcomes for combinations
of actions across time.  Clearly, the state space could be huge for
any significant (time) depth of [re]action.

For these `objective functions', the terrorists might be seeking a
visible breakdown of democratic process by the government (internally
illegal actions) while the governments might be seeking to `shake up'
the terrorists so that they make mistakes out of desperation.
Studying the psychological dynamics of asymetric populations could be
interesting and useful too.

Obviously, coming up with a taxonomy actions to implement requires a
fair amount of imagination and a clear sense of desirable outcomes
for each side.  Actually, it might be as hard to define that as anything.
(e.g. both sides getting caught up in their respective vendettas independent
of global objectives).

The effort would be daunting for isolated academics, but might well be
fruitful to planners given a sufficiently large and capable modeling
and technical staff.  

RR> I don't understand, however whether Swarm has a repository of
RR> simulations that are available and coded in a way that allows them
RR> to seamlessly interact with others in a fedaration.  Is this the
RR> case?  Or is there a way for simulations to interact and share
RR> information about each other without a common language?  If so,
RR> how does it work?

The idea with Swarm is you can define independent agents that perceive
the world in their own way and that can plan complex actions
independent of other agents.  Swarm keeps track of the lowest-common
denominator of time for all the agents and threads together the
actions to call as time moves forward.  This lets the researcher have
a precise top-level view of things that are happening across time and
space.  It's unlike HLA in the sense that the models that it is
designed for tend to have describe small moving parts (that are
implemented with Swarm scheduling) instead of large opaque interfaces
to other legacy model implementations (managed by HLA syncrhonization).

Swarm is a library that provides multi-scale scheduling for agents,
tools to look the the internal state of agents in real time, and
several kinds of landscapes for agents.  Since it is a library you can
use it with other kinds of software.  There is currently well-tested
support for C and Java, and some new support for COM as well.  (It
could be adapted for CORBA (e.g. HLA) without much effort.)  The COM
stuff was done to integrate Swarm with Mozilla (Netscape).  Mozilla
provides a link to very rich XML information representation libraries
which could be used to share simulation metadata with other systems
over the internet.

Please e-mail address@hidden if you'd like to find modelers to hire
to do this kind of work..


                  ==================================
   Swarm-Modelling is for discussion of Simulation and Modelling techniques
   esp. using Swarm.  For list administration needs (esp. [un]subscribing),
   please send a message to <address@hidden> with "help" in the
   body of the message.
                  ==================================