[Battlemesh] ns-3 - Network Simulator

Paul Fuxjaeger fuxjaeger at ftw.at
Wed Nov 18 14:54:04 UTC 2015


On 17.11.15 20:12, fboehm wrote:
> The simulation would also need to include different antenna types 
> that are used at each node (omni, sector, directional). Without 
> having ever simulated a wireless network it sounds like a complex 
> setup to me. The results could be simply bogus because the simulation
> model isn't comprehensive.

IIRC, ns-3 supports directional antenna models:
Input is node positions, antenna orientations and gain characteristics.
Output is additional loss/gain factors which are added to the pairwise
path losses.

You don't seem to need a frequency selective model setup.

So, the "only" thing you have to take care of is to check if the
receiver (and thus interference) model and the overall path loss model
fit to your scenario well enough. Our measurements indicate that the
ns-3 interference model seems to be good enough for most needs (at least
when we are talking about AWGN and Atheros chipsets). Which leaves an
unmatched path loss model to be the main source of prediction error.

> Maybe an acceptable alternative approach is to statically calculate
> a table that stores the level and variance of interference each node
>  generates to every other node. The data in this table would be the 
> input to a network emulator that emulates throughput, latency, 
> jitter, loss.

Calculation of these numbers needs again a well matched path loss model.
But if you can *measure* pairwise path attenuations in your mesh that
could be then turned into a static path loss matrix that can be plugged
into ns-3 as is.

-paul


More information about the Battlemesh mailing list