Re: Combining ABM and CFD in Swarm?

[M Lang / S Railsback]

Uffe Høgsbro Thygesen wrote:
> 
> Hi Swarmers
> 
> I am curious if anyone has used Swarm in applications similar to the one I
> work with. I just downloaded Swarm but I am not really a Swarm user (yet).
...

> My question is: Do you know of anyone who has combined Swarm with a CFD
> package? Or extended Swarm with specialised features for CFD?
> 
> Any pointers are much appreciated. Cheers,
> 
> Uffe

I admire your ambition very much!

We have built several fish models in Swarm
(http://math.humboldt.edu/~simsys/ ), but so far have avoided direct
mixing of fluid mechanics and Swarm. Our trout model uses a
two-dimensional space discretized into irregular rectangular cells; for
each cell we import lookup tables of depth and velocity as a function of
river flow. Then on each time step, each cell interpolates its depth and
velocity from the current flow.

For a model of salmon smolts migrating downstream, we used one set of
node values from a 2-d CFD model as a steady-state representation of the
velocity field. The model uses a discrete (but high resolution) space-
Each time step, the fish search the space for the nearest hydraulic
model node and get their velocity components from that node.

We are also starting to use two-dimensional hydrodynamic models to
generate input to the Swarm model. For example, we could export results
from a CFD model to a geographic information system (e.g., a map of the
river, with depth, velocity, and maybe some other habitat variables
known at each of the CFD nodes); then have the GIS calculate the cell
boundaries so that variation within cells is minimized and variation
among cells is maximized.

For a new model of river backwaters that occasionally experience
tide-like flows, we have also considered putting the hydrodynamic
equations into one of Swarm's two-dimensional square grid space objects.
For us, this has not been too desirable because the hydrodynamic
equations need to be solved at a much shorter time steps than is
desirable for the biological agents. 

In general the Swarm approach of each agent "knowing" where it is in
space (by knowing either a set of coordinates in continuous space, or a
specific cell or node in discrete space) and getting its hydraulic
information from the space, should be good for your needs. The trick
will be getting the hydraulic simulation results where the agents can
find them- perhaps by running the hydraulic model separately and having
the Swarm model read its output file (Swarm can use the HDF5 database
format, which may be an efficient way to transfer the information), or
find a way to link the Swarm model to the hydrodynamic model, or putting
the CFD equations in Swarm. I don't know of anyone that has done any of
these yet.

I'm very curious about your application- could you please send me an
email at address@hidden and tell me more?

We also have written several substantial pieces of code that may be of
help if you are modeling fish.

Steve Railsback
-- 
address@hidden
Lang, Railsback & Assoc.
250 California Ave., Arcata CA 95521
707-822-0453; Fax 822-1868

Until July 1, 2001: 
        Los Alamos, NM
        505-661-4258

                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.