AI-Assisted Dynamic Port and Waveform Switching for Enhancing UL Coverage in 5G NR

Area of Science:

• Telecommunications Engineering
• Wireless Communication Systems
• Machine Learning Applications

Background:

• 5G networks utilize cyclic prefix orthogonal frequency division multiplexing (CP-OFDM) and discrete Fourier transform spread OFDM (DFT-S-OFDM) waveforms.
• Waveform selection impacts power amplifier (PA) efficiency and user throughput, particularly for cell-interior versus cell-edge users.
• Existing static signal-to-noise ratio (SNR) thresholds are insufficient due to dynamic network conditions.

Purpose of the Study:

• To develop an adaptive waveform-switching mechanism for 5G uplink.
• To optimize throughput for both cell-average and cell-edge users.
• To address limitations of static threshold-based waveform selection.

Main Methods:

• Implementation of a deep reinforcement learning (DRL) agent for intelligent waveform switching.
• Optimization of a function based on real-network throughput percentiles, weighted for fairness.
• Utilization of aggregated signal-to-noise ratio (SNR) and timing advance (TA) measurements.
• Inclusion of switching costs (communication interruption) in the DRL model.

Main Results:

• The proposed DRL scheme achieves significant throughput gains for cell-edge users.
• Average network throughput is maintained without degradation.
• The adaptive mechanism effectively handles varying user and channel dynamics.

Conclusions:

• Deep reinforcement learning offers an effective solution for adaptive waveform selection in 5G networks.
• The proposed method enhances user experience, particularly for challenging cell-edge scenarios.
• Considering switching costs improves the practical applicability of waveform selection strategies.