A control theoretic approach to achieve proportional fairness in 802.11e EDCA WLANs

Xiaomin Chen, Ibukunoluwa Akinyemi, Shuang-Hua Yang

Abstract

Researchers from Loughborough University propose a centralized control system using multivariable state-space theory to achieve proportional fairness in 802.11e (Wi-Fi) networks. By dynamically tuning contention windows via a Linear Quadratic Integral (LQI) controller, the system ensures equitable bandwidth distribution across various traffic types while adhering to specific delay requirements.

Introduction

In modern Wi-Fi networks utilizing the 802.11e Enhanced Distributed Channel Access (EDCA) standard, traffic is prioritized into Access Categories (ACs) such as Voice, Video, Best Effort, and Background. While this prioritization supports high-quality real-time services, static default settings often lead to "starvation" of lower-priority traffic in congested environments. This paper addresses the need for a mathematical definition of fairness and an automated mechanism to enforce it under changing network conditions.

Network Model and Problem Formulation

The study considers a network where an Access Point (AP) serves multiple stations (STAs) grouped into different ACs.

Network architecture with an Access Point and multiple Access Categories

The primary objective is to achieve Proportional Fairness, which mathematically maximizes the sum of the logarithms of station throughputs. This is formulated as an optimization problem to determine the ideal transmission attempt probabilities for each station, ensuring fairness while maintaining average delays below critical thresholds (e.g., 250µs for video).

Proposed Methodology: The Control System

To handle the dynamic nature of real-world networks, the authors developed a centralized closed-loop control system. The WLAN is modeled as a discrete Multi-Input Multi-Output (MIMO) Linear Time-Invariant (LTI) system.

Closed-loop control system for WLAN parameter adjustment

The core of the system is a Linear Quadratic Integral (LQI) controller. The AP monitors collision rates and uses the LQI algorithm to adjust the stations' Minimum Contention Window ($CW_{min}$) toward the optimal operating point. Stations provide feedback by "piggybacking" transmission statistics onto existing packets, allowing the AP to recalculate observed collision probabilities ($p^{obs}$) every 100ms and update network parameters in real-time.

Key Insights and Results

Evaluation through Matlab-based simulations revealed several critical insights:

  • Counter-Intuitive Fairness: To achieve proportional fairness, "Best Effort" stations may require higher attempt probabilities than "Video" stations because the latter utilize "TXOP bursting" (sending multiple packets per access).
  • Performance: The controller demonstrated fast convergence, returning the network to a fair state within seconds after sudden changes in network size. It successfully maintained video delays below the 250µs target while ensuring data traffic remained viable.
  • Stability: By treating the Wi-Fi channel as a plant in a control system, the authors provide a framework for rigorous stability and convergence proofs.

Conclusion and Future Work

The LQI-based approach offers a robust method for managing network fairness and delay. While it requires centralized control and potential firmware modifications to allow arbitrary $CW_{min}$ values, it significantly outperforms static heuristic methods. Future research may explore self-optimizing controller parameters and adaptation to non-ideal channel conditions involving external interference.

Summary updated

BibTeX
@article{chen2016control,
  author = {Xiaomin Chen and Ibukunoluwa Akinyemi and Shuang-Hua Yang},
  title = {{A control theoretic approach to achieve proportional fairness in 802.11e EDCA WLANs}},
  journal = {Comput. Commun.},
  year = {2016},
  doi = {10.1016/j.comcom.2015.11.002},
}
Figure