Harnessing frequency diversity in wi-fi networks

Apurv Bhartia, Yi-Chao Chen, Swati Rallapalli, Lili QiuMobiCom 2011 · DOI

TL;DR

This paper tackles the underutilized frequency diversity in Wi-Fi networks, where different subcarriers experience varying signal quality. They introduce a suite of cross-layer techniques, including smart symbol interleaving, CSI-informed partial error correction, and MAC-layer Forward Error Correction, to exploit this diversity and significantly boost throughput.

What Problem Does This Paper Address?

Modern Wi-Fi networks, particularly those using Orthogonal Frequency Division Multiplexing (OFDM), transmit data across multiple subcarriers. While OFDM is robust to multipath fading, a key characteristic of wireless channels is "frequency-selective fading," meaning some subcarriers will have a stronger signal than others at any given moment. This variation, or "frequency diversity," is often treated as noise or a nuisance. However, this paper argues that this inherent frequency diversity is significant (up to 10 dB differences in SNR) and, crucially, can be predicted and exploited. The problem is that existing Wi-Fi systems don't effectively leverage this subcarrier-level signal quality variation, leading to suboptimal performance.

Methods

The authors propose a multi-pronged approach to harness frequency diversity:

  • Smart Symbol Interleaving: Instead of standard interleaving that aims to spread out errors, their approach intelligently maps more important data symbols (like packet headers or critical data bits) to subcarriers with higher Signal-to-Noise Ratios (SNRs). They explore several heuristics, including "smart header," "smart data," and "smart header/data," culminating in a "smart iterative search" that refines symbol-to-subcarrier mapping.
  • CSI-Informed Partial FEC Decoding: They leverage Channel State Information (CSI) – specifically the SNR per subcarrier – to improve Forward Error Correction (FEC) decoding. If an FEC group fails standard decoding, they use CSI to identify and recover symbols on subcarriers with sufficiently high SNRs, effectively salvaging partially corrupted data. This is a significant advantage over traditional methods that discard the entire FEC group.
  • MAC-Layer FEC: To further enhance robustness, they introduce a MAC-layer FEC scheme. This allows for differentiated protection levels for symbols based on their predicted error rates (derived from CSI). Good quality subcarriers might receive less protection, while poor quality ones can be bolstered with additional redundancy, optimizing overall throughput.
  • Joint Rate Adaptation: Finally, they integrate these techniques with rate adaptation. Instead of adapting based solely on frame loss rates, their system considers the improved reliability offered by their techniques and the predicted throughput for various data rates and symbol mappings, selecting the optimal combination.

Datasets

The authors collected their own datasets:

  • Static Traces: CSI data collected from five static Wi-Fi links in an office building environment.
  • Mobile Traces: CSI data collected from four mobile Wi-Fi links, where a sender moved at approximately 1.2 kmph.

These traces were collected using Intel Wi-Fi Link 5300 (iwl5300) NICs on the 5 GHz band (channel 36) to avoid interference. They also used USRP (Universal Software Radio Peripheral) for testbed experiments, injecting narrowband interference to simulate varying channel conditions and create frequency diversity.

Key Insights and Ideas

  • Frequency Diversity is Significant and Predictable: The core insight is that Wi-Fi channels exhibit substantial and predictable frequency diversity. This isn't just random noise; it's a structured variation that can be measured and acted upon.
  • CSI is a Powerful, Accessible Tool: The paper demonstrates that CSI, available on commodity hardware, is a valuable resource for optimizing wireless communication, going beyond simple SNR measurements.
  • Cross-Layer Synergy is Key: The most impactful results come from combining techniques across different layers (PHY, MAC, and application-aware interleaving). No single technique provides the dramatic gains seen when they are integrated.
  • Partial Recovery is More Than Just Error Detection: Using CSI to salvage partially decoded FEC groups significantly boosts throughput by avoiding complete packet loss when only a few symbols are corrupted.
  • MAC-Layer FEC Offers Granular Control: The MAC-layer FEC provides a flexible way to add protection tailored to specific subcarrier quality, which is more adaptable than fixed PHY-layer FEC options.

Key Results

The paper reports significant throughput improvements across various scenarios:

  • Smart Symbol Mapping: Improved throughput by 1.2x to 53.6x compared to standard mapping in static traces.
  • CSI-Based Hints (Partial Recovery): Improved throughput by 1.6x to 5.3x in static traces.
  • MAC-Layer FEC: Improved throughput by 4% to 297% in static traces.
  • Joint Optimization: Combined techniques achieved throughput gains of 2.2x to 4.9x in static traces, and even higher in mobile scenarios.
  • Testbed Experiments: USRP experiments showed improvements of 33-147% for the combined approach over the default scheme.

Limitations and Future Work

  • CSI Prediction Accuracy: While CSI is shown to be temporally stable, high mobility can still degrade prediction accuracy, posing a challenge for robust optimization.
  • Hardware Constraints: The USRP experiments were limited to 800 KHz, which resulted in more homogeneous channels than typical 20-MHz Wi-Fi channels. Future work will explore wider-band SDRs.
  • Robustness to Prediction Errors: The authors plan to explore robust optimization techniques to mitigate the impact of prediction errors, especially in highly mobile environments.
  • Extensive Scenarios: They aim to collect more traces and understand the benefits of their approaches in a wider range of scenarios.
  • Integration with Existing Systems: The paper suggests that their techniques are complementary to existing scheduling algorithms (like those in LTE) and could be integrated with frequency-aware modulation at the PHY layer.

Summary updated

BibTeX
@inproceedings{bhartia2011harnessing,
  author = {Apurv Bhartia and Yi-Chao Chen and Swati Rallapalli and Lili Qiu},
  title = {{Harnessing frequency diversity in wi-fi networks}},
  booktitle = {MobiCom},
  year = {2011},
  doi = {10.1145/2030613.2030642},
}
Figure