Visible Light Communications-Based Indoor Positioning via Compressed Sensing

Kristina Gligoric, Manisha Ajmani, Dejan Vukobratovic, Sinan SinanovicIEEE Communications Letters 2018 · OpenReview

TL;DR

This paper proposes a novel indoor positioning system using Visible Light Communication (VLC) that leverages Compressed Sensing (CS) to efficiently identify nearby LED transmitters. By treating signal separation as a CS problem, the system can accurately estimate a user's location even with overlapping light beams, achieving performance close to theoretical limits.

What Problem Does This Paper Address?

Global Positioning Systems (GPS) are ineffective indoors due to signal obstruction by buildings. Existing indoor positioning systems (IPS) like those using ultrasound, RFID, or Wi-Fi often require additional infrastructure, increasing costs and complexity. Visible Light Communication (VLC), utilizing existing LED lighting, offers a promising alternative due to its high bandwidth, accuracy, and license-free operation. However, accurately determining a user's position in a VLC system, especially when multiple LED light beams overlap, presents a significant challenge. This paper aims to address this by developing a robust and efficient indoor positioning method using VLC and Compressed Sensing.

Methods

The core of the proposed method lies in framing the problem of identifying which LEDs are "visible" to a user device (UD) as a Compressed Sensing (CS) problem.

  1. Signal Model: Each LED transmits its unique binary ID sequence. The UD, equipped with a photodiode, receives a superposition of signals from multiple LEDs. This received signal y can be modeled as y = Sx + n, where S is a matrix containing all LED ID sequences, x represents the product of channel gains and a binary vector indicating LED coverage, and n is noise.
  2. Compressed Sensing for Signal Separation: The key insight is that at any given location, a UD is typically covered by only a small subset of the total LEDs. This makes the vector x "sparse" (having few non-zero elements). CS techniques, specifically Orthogonal Matching Pursuit (OMP) in this paper, are used to efficiently recover this sparse vector x from the received signal y and the known matrix S. This step effectively identifies the set of nearby LEDs and estimates their channel gains.
  3. Proximity-Based Position Estimation: Once the set of detected LEDs (Ŝa) is identified, a simplified proximity method is used for position estimation. The UD's position is estimated as the centroid of the union of the coverage areas of these detected LEDs. This method is chosen for its low complexity and its ability to leverage the increased number of reference points provided by overlapping light beams.

Datasets

The paper does not explicitly mention using pre-existing datasets. Instead, it describes a simulated, large-scale indoor open-plan office space scenario for evaluation. This scenario involves:

  • A square-shaped floor area (50m x 50m) with a ceiling height of 3m.
  • LEDs arranged on an equidistant grid on the ceiling.
  • A user device (UD) equipped with a photodiode (PD) placed at an unknown location.
  • Random distribution of 1000 UDs for each simulation run.
  • The simulation parameters include varying Signal-to-Noise Ratio (SNR), LED ID signature length (M), and LED coverage radius (r).

Key Insights and Ideas

  • VLC as a Positioning Tool: The paper reinforces the viability of using existing LED lighting infrastructure for accurate indoor positioning, reducing the need for dedicated hardware.
  • Compressed Sensing for Efficiency: Applying CS to the signal separation problem is a crucial innovation. It allows the system to efficiently identify relevant LED signals from a potentially large number of transmitters, even with overlapping beams, without needing to decode every single LED.
  • Overlapping Beams Enhance Accuracy: The authors highlight that overlapping light beam regions, which might seem problematic, actually improve positioning accuracy by increasing the number of detected reference points.
  • Proximity Method Sufficiency: For position estimation after signal separation, a simplified proximity method proves effective, demonstrating that complex geometric calculations might not always be necessary for good performance.

Key Results

The paper presents several key findings from their simulations:

  • SNR Impact: As the Signal-to-Noise Ratio (SNR) increases, both the Mean Position Error (MPE) and the Support Recovery Error (SRE) decrease, approaching minimal achievable values. For SNR > 35dB, performance stabilizes.
  • ID Length (M): The Mean Position Error (MPE) is relatively insensitive to the LED ID signature length (M) beyond a certain point. For instance, M=75 bits was found to be sufficient with minimal loss in MPE and only a slight loss in SRE, demonstrating efficiency.
  • Coverage Radius (r): The relationship between coverage radius and positioning accuracy is non-monotone. The proposed method can match the minimum MPE of the proximity method for certain parameters, and increasing SNR is suggested for better performance with larger coverage radii.
  • LED Density: Increasing LED density (NLED) generally leads to improved positioning accuracy, with performance shifting towards the minimum MPE bound as SNR increases.
  • Accuracy: The proposed CS-based method's MPE gradually approaches the lower bound of the proximity method, indicating high accuracy. For example, with a fixed geometry and SNR of 20dB, the MPE can be around 0.4m.

Limitations and Future Work

  • Assumptions: The model assumes line-of-sight (LOS) communication and ignores signal reflections. It also assumes perfect synchronization of LED transmissions and a constant channel gain during the transmission of an ID sequence.
  • Proximity Method Simplification: The proximity method used for position estimation effectively disregards channel gains, relying solely on the detected set of LEDs. While effective, this simplification might limit performance in scenarios where channel gain variations are significant.
  • Future Work: The paper suggests that for larger coverage radii, increasing SNR would be necessary to maintain accuracy. Further exploration into optimizing the CS recovery algorithms and potentially incorporating channel gain information into the proximity method could be avenues for future research. The authors also note the "group-sparsity" property of the LED ID detection problem, hinting at potential for more advanced CS techniques.

Summary updated

Abstract

This letter presents an approach for visible light communication-based indoor positioning using compressed sensing. We consider a large number of light emitting diodes (LEDs) simultaneously transmitting their positional information and a user device equipped with a photo diode. By casting the LED signal separation problem into an equivalent compressed sensing framework, the user device is able to detect the set of nearby LEDs using sparse signal recovery algorithms. From this set and using proximity method, position estimation is proposed based on the concept that if signal separation is possible, then the overlapping light beam regions lead to decrease in positioning error due to increase in the number of reference points. The proposed method is evaluated in a LED-illuminated large-scale indoor open-plan office space scenario. The positioning accuracy is compared against the positioning error lower bound of the proximity method, for various system parameters.
BibTeX
@article{gligoric2018visible,
  author = {Kristina Gligoric and Manisha Ajmani and Dejan Vukobratovic and Sinan Sinanovic},
  title = {{Visible Light Communications-Based Indoor Positioning via Compressed Sensing}},
  journal = {IEEE Communications Letters},
  year = {2018},
  url = {https://openreview.net/forum?id=Z4q06EcMdS},
}
Figure