[GSoC 2026] Review request - NR energy consumption framework (cttc-lena/nr !408)

10 views
Skip to first unread message

Nipuna Dulara

unread,
Jul 29, 2026, 12:56:34 AMJul 29
to ns-dev...@googlegroups.com

Hello ns-developers,

I'm Nipuna Dulara, working this GSoC on an energy consumption framework for the 5G-LENA NR module, mentored by the CTTC OpenSim team (Biljana, Katerina and Joao). I'd like to ask for review of the first merge request, which is now ready.

About the project

5G-LENA models NR throughput and latency in detail, but until now it had no way to answer "how much power does this cost?". The project adds a standards-based energy layer so users can do per-device energy accounting and 3GPP power-saving evaluation directly in ns-3:

  • gNB power model implementing TR 38.864 (P1–P5 relative power levels keyed by BS category/reference set, the DL/UL active-power formulas with the sa/sf/sp activity factors and PA-efficiency modes, and symbol-level slot energy).
  • UE power model implementing TR 38.840 (FR1/FR2 relative-power states, the BWP/antenna/blind-decoding scalings, UL power interpolation, occupancy accounting) plus a connected mode DRX power model (onDuration / inactivity / long-cycle timers driving light/deep sleep).
  • PHY-side listeners that drive the models from existing PHY traces, so the models stay pure and a scenario doesn't have to be rewritten. Both models plug into the standard ns3::energy framework as DeviceEnergyModels.

The models are validated against real hardware power traces (base-station and UE), matching measured shape and averages closely.

What to review now

  • !408 ( energy models and listeners ). It contains the two power models, the DRX model, the two PHY listeners, and the test suites. It also includes one small, additive core-PHY change worth your attention: a new SlotEnergyStats trace source on NrGnbPhy (~15 lines in GenerateAllocationStatistics) reporting the slot's used symbols split by direction (DL/UL data and control, with DL REG counts). The existing SlotDataStats/SlotCtrlStats traces are untouched; this is what lets the gNB charge UL reception at P_UL and PDCCH at P_DL rather than billing everything as DL/idle.  https://gitlab.com/cttc-lena/nr/-/merge_requests/408

The remaining two MRs are still in draft and i'd like to invite you all to review them as well:

Thanks very much for your time. Feedback on the design (especially the core-PHY trace and the 3GPP mapping) is very welcome.


Best regards, Nipuna Dulara University of Moratuwa GSoC 2026 - ns-3 / 5G-LENA (CTTC)

Reply all
Reply to author
Forward
0 new messages