[Ns-developers] PropagationLossModel behaviour
tjkopena at cs.drexel.edu
Tue Jun 3 10:21:22 PDT 2008
On Mon, Jun 2, 2008 at 6:52 PM, Hagen Paul Pfeifer <hagen at jauu.net> wrote:
> If we want to visualize or do some other analysis and get this value at time
> n. The result will differ from n because we call the method a second time.
> So there is a mismatch between (albeit the delta is small) the value that the
> channel used and the public[tm] one. IMHO the PropagationLossModel should be
> correlated to the current state (e.g. time). Or did I missed something (I mean
> it is a lot of code and interaction between several subparts of ns-3)? ;)
I proceed w/ a disclaimer that I am neither an RF nor a
Connecting all models that might be visualized to time in a
deterministic way (even given the RNG seed) sounds difficult. Are you
aware of models that would operate in such a fashion? Any models I
can think of would basically amount to storing the value for the given
time after first computation. How would you deal w/ values generated
at the same simulation time (definitely possible) but for different
events? Discriminating the events sounds messy.
Like Lacage said, I think if you want to visualize the exact
propagation variables at that time you need to add trace sources if
they're not there and connect them to some sort of log.
Alternatively, you could get away with something approximate. For
example: Sometime soon I need to produce some network state
visualizations for a paper, most likely using Mathieu's WiFi model.
I'm leaning toward plotting the radio ranges via some mechanism like
plotting circles at the distance w/ 95% chance of successful delivery.
Note that this is more reasonable for a homogenous environment, as
I'm using currently, where the receiver stats in some sense don't
matter. If nodes could have different receivers then I would probably
try and plot expected distance for a given power level, but the former
I would be interested to hear comments you or anyone else has.
- joe kopena
right here and now
More information about the Ns-developers