Researchers propose using a Spectral Temporal Graph Neural Network (StemGNN) to predict future user equipment scheduling states, addressing the performance degradation in distributed 5G networks caused by backhaul latency. By replacing stale inter-cell information with these predictions, the framework mitigates the negative effects of delayed coordination on coordinated beamforming.

  • StemGNN achieves a mean scheduling prediction accuracy of 87.57%, outperforming LSTM, GRU, Simple RNN, and Markov chain baselines across all evaluated horizons.
  • The model shows gains of up to 7.71% over LSTM at longer horizons where inter-user equipment structural dependencies dominate.
  • Integrated into coordinated beamforming, the predictions recover 57-73% of the sum rate loss caused by one transmission time interval of backhaul delay.
  • This approach improves sum rate by 9.58-14.35% over the no-prediction baseline and recovers up to 83% of the Lag-1 fairness loss for cell-edge users.

Treating backhaul latency as a spatio-temporal forecasting problem provides an effective method for maintaining robust inter-cell coordination in delay-constrained networks.