Heterogeneous spectrum sensing: challenges and methodologies

Wei Liu, Mikołaj Chwalisz, Carolina Fortuna, Eli De Poorter, Jan Hauer, Daan Pareit, Lieven Hollevoet, Ingrid MoermanEURASIP 2015 · DOI

Abstract

Distributed sensing is commonly used to obtain accurate spectral information over a large area. More and more heterogeneous devices are being incorporated in distributed sensing with the aim of obtaining more flexible sensing performance at lower cost. Although the concept of combining the strengths of various sensing devices is promising, the question of how to compare and combine the heterogeneous sensing results in a meaningful way is still open. To this end, this paper proposes a set of methodologies that are derived from several spectrum sensing experiments using heterogeneous sensing solutions. Each of the solutions offers different radio frequency front-end flexibility, sensing speed and accuracy and varies in the way the samples are processed and stored. The proposed methodologies cover four fundamental aspects in heterogeneous sensing: (i) storing experiment descriptions and heterogeneous results in a common data format; (ii) coping with different measurement resolutions (in time or frequency domain); (iii) calibrating devices under strictly controlled conditions and (iv) processing techniques to efficiently analyse the obtained results. We believe that this paper provides an important first step towards a standardized and systematic approach of heterogeneous sensing solutions.

Introduction

To improve spectrum usage, "Cognitive Radios" must detect when a licensed frequency is idle. While high-end equipment can do this easily, it is expensive to deploy at scale. The alternative is distributed sensing using many low-cost, heterogeneous devices, such as sensor nodes or USB dongles. However, these devices present significant coordination challenges. They record data in different formats, have wildly different sampling resolutions, and possess varying noise floors and sensitivities. Without a systematic way to align this data, the resulting "radio environment map" is often inaccurate or impossible to synthesize. This research addresses the lack of a standardized methodology for storing, calibrating, and processing data across diverse sensing hardware.

Methodology

The proposed framework bridges the hardware gap through a four-pillar methodology designed to standardize the sensing workflow.

The high-level methodology flow for heterogeneous spectrum sensing.

1. Common Data Format (CDF)

The authors introduce a machine-readable, device-independent structure based on XML. This Common Data Format (CDF) organizes experiment data into three logical sections: an Abstract defining goals, Meta-information describing hardware and signal parameters, and Iterations containing the actual measurement traces.

The hierarchical structure of the Experiment Description within the Common Data Format.

2. Resolution Alignment

Devices often sample at different rates (e.g., 10 kHz vs. 1 MHz). To enable "apples-to-apples" comparisons, the authors employ a post-processing integration technique. By integrating the Power Spectral Density (PSD) over a common frequency interval, they create a unified metric that accounts for varying resolution bandwidths.

3. Calibration Workflow

To mitigate hardware-specific biases, a "coaxial-cable-based" calibration is proposed. By bypassing the wireless interface and connecting devices directly to a high-end signal generator via splitters and attenuators, researchers can calculate a precise "power offset" for every device, correcting for internal noise and gain variations.

4. Data Mining and Analysis

Advanced techniques like regression and clustering are used to handle large datasets. For instance, clustering can automatically distinguish between Line-of-Sight (LOS) data and shadowed (NLOS) data, improving the accuracy of path loss models.

Experimental Evaluation

The methodology was validated using a diverse fleet of hardware, including the ultra-low-cost TelosB sensor node, commercial Wi-Spy and Airmagnet USB analyzers, the flexible USRP SDR platform, and a custom high-performance imec sensing engine. Experiments were conducted in an indoor cafeteria with 23 measurement locations to test real-world shadowing effects, such as signals blocked by large appliances.

Key Insights and Results

  • The Resolution Trap: Raw traces from devices with coarser resolution bandwidths naturally show higher PSD levels. The proposed integration step is crucial to prevent incorrect conclusions about energy levels.
  • Calibration Necessity: Even high-end SDRs like the USRP showed power offsets of nearly 8 dB compared to reference signals, proving that factory calibration is insufficient for collaborative sensing.
  • Sensitivity Performance: The custom imec sensing engine and Wi-Spy dongle demonstrated the highest sensitivity, while the TelosB and USRP were limited by higher noise floors.
  • Shadowing Analysis: By applying clustering, the authors demonstrated that shadowing (NLOS) fundamentally changes the path loss exponent, allowing for more sophisticated radio environment maps that reflect physical layouts.

Conclusion and Future Work

This work provides a foundational step toward standardized heterogeneous sensing. While the CDF is universal, it currently requires device-specific "wrapper" scripts for data conversion. Future work will focus on incorporating antenna gain patterns into calibration—as the current method uses coaxial cables—and automating validation procedures to simplify the integration of new hardware types.

Summary updated

Abstract

Distributed sensing is commonly used to obtain accurate spectral information over a large area. More and more heterogeneous devices are being incorporated in distributed sensing with the aim of obtaining more flexible sensing performance at lower cost. Although the concept of combining the strengths of various sensing devices is promising, the question of how to compare and combine the heterogeneous sensing results in a meaningful way is still open. To this end, this paper proposes a set of methodologies that are derived from several spectrum sensing experiments using heterogeneous sensing solutions. Each of the solutions offers different radio frequency front-end flexibility, sensing speed and accuracy and varies in the way the samples are processed and stored. The proposed methodologies cover four fundamental aspects in heterogeneous sensing: (i) storing experiment descriptions and heterogeneous results in a common data format; (ii) coping with different measurement resolutions (in time or frequency domain); (iii) calibrating devices under strictly controlled conditions and (iv) processing techniques to efficiently analyse the obtained results. We believe that this paper provides an important first step towards a standardized and systematic approach of heterogeneous sensing solutions.
BibTeX
@inproceedings{liu2015heterogeneous,
  author = {Wei Liu and Mikołaj Chwalisz and Carolina Fortuna and Eli De Poorter and Jan Hauer and Daan Pareit and Lieven Hollevoet and Ingrid Moerman},
  title = {{Heterogeneous spectrum sensing: challenges and methodologies}},
  booktitle = {EURASIP},
  year = {2015},
  doi = {10.1186/s13638-015-0291-8},
}
Figure