Optimizing Simulation Runtime by Precomputing the Channel Matrix for Static UEs

[122g...@gmail.com]

Hello everyone,

I have an topology where the positions of the UEs remain static and do not change.
To reduce runtime as the number of UEs increases, I would like to compute the channel matrix only once, instead of recalculating it throughout the entire simulation.

Do you think this is a good idea?
If so, does anyone have suggestions on how to implement it efficiently?

[Tommaso Pecorella]

I don't know enough of the channel model, but you have to also consider the interference between nodes. That can not be pre-computed :(

[122g...@gmail.com]

Hi Tommaso,

Thanks for the comment!

In my setup, all nodes are static without any time-varying fading. Since the channel matrix is fixed, the interference — which depends on these static channel gains — can also be derived accordingly.

So, I believe precomputing the channel matrix once is valid and can help reduce runtime.

Let me know if you see any issue with this approach.

ב-יום חמישי, 1 במאי 2025 בשעה 16:20:54 UTC-4, Tommaso Pecorella כתב/ה:

[Kent Huns]

I got the same message on the 5G-LENA's forum, but replying here.

Unfortunately, your idea would not be accepted in mobile communication.

You must have seen rain, wood, mirage in hot weather, car or pedestrian. Even if Tx & Rx are fixed, that doesn't mean the channel matrix is.

Especially in 5G context, channel fluctuation shouldn't be ignored because its core technologies like

NLOS propagation, OFDM or SD-MIMO(=intentional multipath transmission) have a lot to do with reflection.

In fact, 5G's typical channel model (= clustered delay line model in 3GPP TR 38.901) uses random number in the calculation every time.

Omitting this process for calculation cost is not optimization but just simplification for poor machines. And we should take the time for a good simulation.