[Paper Review] Recent Results on Proportional Fair Scheduling for mmWave-based Industrial Wireless Networks

Jiteng Ma, Adnan Aijaz+1 more|arXiv (Cornell University)|Jul 11, 2020

Millimeter-Wave Propagation and Modeling16 references4 citations

TL;DR

This paper proposes an Enhanced Proportional Fair (EPF) scheduler for mmWave-based industrial wireless networks to improve fairness and responsiveness in dynamic, scatter-rich environments. By modifying the standard Proportional Fair (SPF) algorithm to prioritize users in non-line-of-sight (NLOS) conditions and accelerate reaction to channel fluctuations, EPF achieves up to 10.7% better 'beyond 95th percentile' latency and 0.994 fairness index in throughput, outperforming SPF in both fairness and robustness under rapid channel variations.

ABSTRACT

Millimeter wave (mmWave) communication has recently attracted significant attention from both industrial and academic communities. The large bandwidth availability as well as low interference nature of mmWave spectrum is particularly attractive for industrial communication. However, inherent challenges such as coverage and blockage of mmWave communication cause highly fluctuated channel quality. This paper explores wireless medium access control (MAC) schedulers for mmWave-based industrial wireless applications. Our objective is to design a high-performance and enhanced fairness MAC scheduling algorithm that responds rapidly to channel variations. The key contribution of our work is a method to modify the standard proportional fair (SPF) scheduler. It introduces more flexibility and dynamic properties. Compared to the SPF, our enhanced proportional fair (EPF) scheduler not only improves the priority for users in poor channel conditions but also accelerates the reaction time in fluctuated channel conditions. By providing higher fairness for all users and enhancing system robustness, it particularly adapts to the scatter-rich industrial mmWave communication environment. Through extensive performance evaluation based on the widely accepted network simulator (ns-3), we show that the new scheduler achieves better performance in terms of delivering ultra-low latency and reliable services over mmWave-based industrial communication.

Motivation & Objective

• To address the limitations of standard Proportional Fair (SPF) scheduling in rapidly fluctuating mmWave channels common in industrial environments.
• To improve fairness for users in non-line-of-sight (NLOS) conditions, which suffer from deep fades and blockage in mmWave communications.
• To reduce queuing delay and improve ultra-low latency performance in mmWave-based industrial IoT applications.
• To design a MAC-layer scheduler that dynamically adapts to channel variations while maintaining high system throughput.
• To evaluate the performance of the proposed scheduler under realistic industrial scenarios with high mobility and dense scattering.

Proposed method

• The EPF scheduler modifies the standard SPF algorithm by introducing a dynamic weighting factor that increases the priority of users experiencing poor channel conditions, especially in NLOS scenarios.
• It accelerates reaction to channel fluctuations by adjusting the fairness metric to respond more quickly to changes in channel quality, reducing queuing delay.
• The scheduler uses a modified throughput-to-fairness trade-off function that emphasizes user fairness in poor channel states while maintaining overall system efficiency.
• The algorithm is implemented and evaluated using the mmWave communication module in the ns-3 network simulator with realistic 3GPP statistical channel models.
• Flexible TTI and adaptive modulation and coding are used to reflect real-world mmWave system constraints.
• Performance is evaluated using Jain’s fairness index for throughput and latency, and 'beyond 95th percentile' latency to quantify worst-case delay performance.

Experimental results

Research questions

• RQ1 How does the EPF scheduler improve fairness for users in non-line-of-sight (NLOS) conditions compared to the standard Proportional Fair (SPF) scheduler?
• RQ2 To what extent does the EPF scheduler reduce worst-case (95th percentile) latency in rapidly fading mmWave channels?
• RQ3 What is the trade-off between system throughput and fairness when prioritizing NLOS users in scatter-rich industrial environments?
• RQ4 How does the EPF scheduler respond to rapid channel variations compared to SPF, especially during LOS-to-NLOS transitions?
• RQ5 Can the EPF scheduler maintain high reliability and ultra-low latency in mmWave-based industrial wireless networks under realistic traffic and mobility models?

Key findings

• The EPF scheduler achieves a throughput fairness index of 0.994, significantly improving fairness compared to SPF in NLOS-heavy environments.
• In Case 1 (NLOS mobility), the EPF scheduler reduces 'beyond 95th percentile' latency by 10.7% compared to SPF, indicating faster adaptation to channel changes.
• In Case 2 (random mobility and scattering), the 'beyond 95th percentile' latency improves by 12.3% with EPF, demonstrating robustness in realistic conditions.
• The EPF scheduler enhances NLOS user throughput and latency performance, even at the cost of slight degradation in LOS user performance.
• The EPF scheduler improves latency fairness by 2.2% compared to SPF, as measured by the Jain’s fairness index in Case 2.
• The proposed EPF scheduler effectively balances system performance and fairness, making it suitable for ultra-reliable, low-latency industrial mmWave applications.