[Ns-developers] Reproducing Broch AODV Performance Numbers

[Kevin Peters]

On Thu, Apr 15, 2010 at 4:44 AM, Nicola Baldo <nbaldo at cttc.es> wrote:

> Is is possible to make a simple change to ns-3 such that it
>> receives based on received signal strength like ns-2 in order to
>> rule this out or in?  Just a temporary toggle to check if this
>> could explain the differences in performance?
>>
>
> As a quick hack, you can modify YansWifiPhy::EndReceive as follows:
>
> - if (m_random.GetValue () > snrPer.per)
> + if (event->GetRxPowerW () > RxPowerThresholdW)
>
> this bypasses the interference model of YansWifiPhy, and allows you to use
> a simple threshold, as ns-2 does.
>
> As for the value of the CS threshold in the Broch paper, you should sett
> the EnergyDetectionThreshold attribute in YansWifiPhy to an equivalent
> value.
>
> As for capture... that's not trivial. Maybe just neglect it for the moment
> and see what results you get.
>
> Finally, let me say that if anybody wants to develop a new WifiPhy
> implementation that uses a simple protocol interference model (i.e., a
> threshold model like the one in ns-2), I would support merging that.
> But I think it should be separate from YansWifiPhy.
>
> Nicola
>

Here is what I tried.  From trial and error testing in default ns-3 to find
the optimal transmit power for various distances between a single sender and
receiver, I found the receive power (Pr) for the Friis propagation loss
model to be constant at -78.832 dBm.  I converted this to Watts and made the
following change to YansWifiPhy:

double RxPowerThresholdW = 0.000000000013085791612451009; //
1.3085791612451009e-11
if (event->GetRxPowerW () > RxPowerThresholdW)

I didn't see any specific CS threshold numbers in the Broch paper, so I just
went with the ns-3 default of -96.

With this, through trial and error with one sender and receiver 250 m apart,
I found the optimal transmission power to be 13.9.

Does this sound like a correct approach to simplifying the physical layer to
be like ns-2 with the receive threshold?

So, I tested the above configuration with the AODV Broch scenario and
default ns-3 AODV and it seems to have made things even worse.  After a
couple hundred seconds in the simulation, the performance degrades to 0%
PDR.  It started with decent performance but quickly approached 0% PDR after
about 100 simulation seconds.

Thanks,
Kevin