[Paper Review] Fairness-Aware Scheduling in Multi-Numerology Based 5G New Radio

Ahmet Yazar, Hüseyin Arslan|arXiv (Cornell University)|Jun 11, 2018

PAPR reduction in OFDM12 references3 citations

TL;DR

This paper proposes two fairness-aware scheduling algorithms for multi-numerology 5G New Radio (NR) systems to mitigate inter-numerology interference (INI) and improve fairness for edge users. By strategically scheduling users based on power levels and subcarrier positions, the algorithms reduce SIR variance and enhance fairness without sacrificing spectral efficiency, achieving up to 10 dB SIR improvement for edge UEs compared to random scheduling.

ABSTRACT

Multi-numerology waveform based 5G New Radio (NR) systems offer great flexibility for different requirements of users and services. Providing fairness between users is not an easy task due to inter-numerology interference (INI) between multiple numerologies. This paper proposes two novel scheduling algorithms to provide fairness for all users, especially at the edges of numerologies. Signal-to-noise ratio (SIR) results for multi-numerology systems are obtained through computer simulations.

Motivation & Objective

• Address fairness degradation in multi-numerology 5G NR due to inter-numerology interference (INI) and power offset effects at numerology edges.
• Mitigate signal-to-interference ratio (SIR) degradation for edge users caused by high power offsets and INI from adjacent numerologies.
• Develop implementable scheduling algorithms that enhance fairness while maintaining spectral efficiency under fixed guard bands.
• Balance fairness for both edge and inner users across multiple numerologies with varying subcarrier spacing.
• Enable practical deployment within 3GPP standards through adaptive, low-complexity scheduling strategies.

Proposed method

• Propose two novel scheduling algorithms that prioritize edge users of different numerologies based on power level and subcarrier position to minimize INI impact.
• Use a power offset-based scheduling strategy that avoids assigning high power levels to edge users of one numerology when they interfere with edge users of another.
• Apply Monte Carlo simulation with 1,000 iterations to estimate average SIR and INI across subcarriers under varying power offset conditions (0–10 dB).
• Employ fast Fourier transform (FFT) with N_ref = 4096 points and CP ratio of 1/16 for accurate SIR and INI estimation in OFDM-based multi-numerology systems.
• Compare proposed algorithms against random scheduling using cumulative distribution function (CDF) curves of SIR for edge and inner users.
• Optimize scheduling by minimizing SIR variance across users to ensure stable performance and fairness over time.

Experimental results

Research questions

• RQ1 How does inter-numerology interference (INI) affect the fairness of edge users in multi-numerology 5G NR systems?
• RQ2 What is the impact of power offset on SIR degradation for edge users across different numerologies?
• RQ3 Can scheduling strategies reduce SIR variance and improve fairness without increasing guard band requirements?
• RQ4 How do inner and edge users of different numerologies respond to power-level variations in the presence of INI?
• RQ5 To what extent can fairness-aware scheduling outperform random scheduling in multi-numerology 5G NR?

Key findings

• In random scheduling, edge users in the same numerology experience at least 7 dB SIR difference, indicating inherent unfairness.
• A 3 dB power offset on a NUM-2 edge UE causes a 5.7 dB SIR drop in the NUM-1 edge UE, demonstrating severe interference impact.
• A 3 dB power offset on a NUM-2 inner UE causes only a 2.8 dB SIR drop in the NUM-1 edge UE, showing less cross-interference than edge-to-edge power increases.
• Symmetric power increases across both numerologies result in a small SIR improvement for edge UEs but increase interference on inner UEs proportionally.
• The proposed fairness-aware algorithms reduce SIR variance significantly compared to random scheduling, as shown by flatter CDF curves for edge users.
• Algorithm 2 slightly improves fairness for inner users compared to Algorithm 1, especially when edge UEs have wider bandwidths, due to reduced side lobe leakage.