[Repast-developer] Developing RePast to behave as an HLA federate

[Rob M. <msc...@cs...>]

Hi all,

Some may be interested in my project here at the University of Birmingham,
UK.

I am creating a package that will (hopefully) enable RePast models to
behave as HLA federates using the IEEE 1516 standard.

For those not familiar the HLA standard is a DoD sponsored framework for
simulation systems interoperability, it defines an abstract object model
and synchronisation semantics which are used by federates (individual
simulations) interacting within a federation (a group of federates
communicating through a shared 'runtime' or RTI).

An HLA federation is intrinsically distributed, allowing parallel
computation of individual federate-state whilst providing a library for
the synchronisation of these states.

My plans at the moment are to develop a package whose primary aims are:

-To allow RePast models to interact with other models (maybe RePast maybe
otherwise) through a shared object model and an RTI.
-To allow a RePast model to be split in to seperate component sections
(a space and a collection of agents for example) which can then be
modelled in seperate federates (RePast instances) achieveing greater
parallelism and performance gains.

My current architecture looks like this:

-A number of classes which extend critical RePast components
(ScheduleBase, SimInit, etc...), adding the functionality and behavioural
constraints demanded of an HLA federate

-A number of library methods for serialising RePast objects and their
attributes, which can then be passed through an HLA RTI to other federates

-A number of interfaces which HLA compliant RePast models can implement

-A 'front-end' like component that handles most low-level interaction
between a RePast federate and the associated RTI, the front-end's most
arduous task will be the initialisation of the federation (discovery,
joining, object-model arbitration, etc...) although it will also handle
time-advance constraints and call-back notification as the federation
makes calls to the model with notifications of event.

If any of this interests anyone I will be happy to keep them informed on a
personal basis of how things are developing and would be delighted to
supply them with the eventual framework and sourcecode.

You can find out more about the HLA at:

www.dmso.mil/public/transition/hla/

The only major weakness at present is the relatively high expense of HLA
RTI implementations (with the sole implementation of IEEE 1516 costing at
least $6000), but free (low-capacity) versions are still out there (again
I will be happy to discuss over personal e-mails).

Rob.