Beam ID Coupling with The Scheduling Decisions

[Amirhassan Babazadeh Darabi]

Hello, 

I am simulating beamforming and beam tracking in mmWave with the 5G-LENA module, and I have hit some interesting issues I wanted to discuss. I have a simulation scenario where a UE rotates around the gNB, and I use two beamforming methods: DirectPathBeamforming and CellScanBeamforming. Everything works fine with the DirectPathBeamforming because no beamID is associated with the beamforming vectors (all have {0, 0} beamIDs). However, in the CellScanBeamforming, beamIDs are different and change when the beamforming vectors change, and because the scheduling part is coupled with the beamID concept, it raises an error, which is attached along with the code. I was wondering if I am doing something wrong, or if it is the limitation of 5G-LENA since the separation of active users is done based on beamID.

I am simulating on ns3.47 and the latest version of 5G-LENA (4.2).

[Gabriel Ferreira]

This seems like the user beam changed during a HARQ retransmission.
This may be because of RLF, or we just forgot to swap that beam for the HARQ when the beam associated with the device changed.
I don't know for sure. We need to check.

[Amirhassan Babazadeh Darabi]

If we seed the channel in both cases so we have the same channel in each simulation (DirectPath and CellScan), we should experience RLF in both, which we don't, so it isn't an RLF issue. Maybe the second one is true. I will check whether a retransmission is occurring while we are changing the beam and whether the HARQ process is updated. Thank you.