Coordination of multi layer scheduling

[Kent Huns]

Hi all,

I hope this topic caught someone's interest.

MU-MIMO has been supported by recent 5G-LENA(v3.0-). 

I read source files and found MAC scheduling done in below functions.

  nr-mac-scheduler-ns3.cc :: DoScheduleDlData()

   --  nr-mac-scheduler-ofdma(tdma).cc :: AssignDLRBG()

   --  nr-mac-scheduler-ofdma(tdma).cc :: CreateDlDci()

This flow doesn't have any index or identifier of MIMO layers,

so I think each layer of MIMO is scheduled independently.

However, in realistic SU & MU-MIMO, some DCI parameters are physically limited.

(and should be limited for avoiding interference.)

I'd like to consider the dependence of MIMO layers.

To do so, I need to memorize all DCIs in the same TTI and forget them at the next TTI.

Maybe their functions are easy to write, but I have a problem about the latter cache clear.

How can I detect the moment when the last layer's CreateDlDci() is finished?

I would be happy if you could share ideas.

Thanks,

Kent

[Biljana B.]

Hi Kent, 

5G-LENA does not support MU-MIMO, but only SU-MIMO. The parameter/variable rank indicates the number of layers in the scheduler code.  So, the underlined question in your email does not apply for the current 5G-LENA code.

We would like in the future to work towards MU-MIMO. Are you working on it?

Thanks,

Biljana

[Kent Huns]

Hi Bilijana,

I'm sorry I misunderstood current feature.

I see now 5G-LENA supports PMI-based SU-MIMO from the release note of v3.0.

If now multi beamforming vectors are managed by PMI, 

does it mean that AssignDLRBG() & CreateDlDci() are called only once in a TTI and

the resource mapping(m_sym, m_rbgBitmask) of each MIMO layer is exactly the same?

MU-MIMO would be feasible at least in Tx scheduling, if gNB could create DCIs for UEs in different BeamIDs.

I think there must be many other things to be modified, but I'd like to confirm how MIMO is scheduled now first.

>> Are you working on it?

I just started thinking about it, and I haven't made any source code.

Best Regards,

Kent

[Biljana Bojovic]

Hi Kent,  please find inline.

On Fri, Sep 27, 2024 at 9:33 PM Kent Huns <kent.sa...@gmail.com> wrote:

Hi Bilijana,

I'm sorry I misunderstood current feature.

I see now 5G-LENA supports PMI-based SU-MIMO from the release note of v3.0.

If now multi beamforming vectors are managed by PMI, 

does it mean that AssignDLRBG() & CreateDlDci() are called only once in a TTI and

the resource mapping(m_sym, m_rbgBitmask) of each MIMO layer is exactly the same?

The scheduling is performed once per slot, and the resource mapping is the same for each MIMO layer.

 

MU-MIMO would be feasible at least in Tx scheduling, if gNB could create DCIs for UEs in different BeamIDs.

In the current implementation, BeamID only refers to the analog beamforming part. But, yes, we would have "overlapping" assignments in the DCI, and at the NrSpectrumPhy of the UE would be able to distinguish the signals based on RNTI, as it is done in the current code. NrInterference is already prepared to correctly distinguish the interference coming from the same cell and the other cells. I would expect the main efforts in the changes related to the implementation of the Type II codebooks, and the scheduler.

 

--
You received this message because you are subscribed to the Google Groups "5G-LENA-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 5g-lena-user...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/5g-lena-users/e4148110-31b1-4622-a05e-aecf4dc93ee0n%40googlegroups.com.

[Kent Huns]

Hi Biljana,

Thank you so much for the detailed answers.

>>we would have "overlapping" assignments in the DCI

It sounds quite a good. I'm also interested in the coverage gap between PDCCH and PDSCH, and this feature makes it clearer.

And I realized my above explanation had missed the point.

I know about BeamID but here I've assumed  "BeamID" = "one of analog BF vectors" = "PDCCH coverage" = "group of UEs they can do OFDMA together".

This is just my assumption and may not be general.

Moreover, I'm not sure whether DCIs must be distributed in each layer or not when Type II codebook.

I'll continue to research about the scheduling of MU-MIMO.

Best regards,

Kent

[Kent Huns]

P.S.

Let me share my image. (It may contain mistakes. And it may not need to be this complicated.)

Thanks,

Kent