Radio Sensing with Large Intelligent Surface for 6G

Cristian J. Vaca-Rubio, Pablo Ramirez-Espinosa, Kimmo Kansanen, Zheng‐Hua Tan, Elisabeth de CarvalhoICASSP 2021

TL;DR

This paper demonstrates how Large Intelligent Surfaces (LIS)—a cornerstone of future 6G networks—can be used as high-resolution "radio cameras" to sense the physical environment. By processing ambient communication signals with computer vision techniques, the researchers achieved nearly 100% accuracy in detecting active devices and 98% accuracy in detecting passive "unconnected" humans, even in noisy environments.

What Problem Does This Paper Address?

In the transition from 5G to 6G, the goal is to move beyond just moving data; we want networks that can "see" and "feel" their surroundings. This is known as Integrated Sensing and Communication (ISAC).

Current sensing methods often require dedicated hardware (like radar) or rely on Channel State Information (CSI) from Wi-Fi, which has limited resolution and often struggles to detect "passive" objects—people or things that aren't carrying a transmitting device. The challenge is: how do we achieve high-resolution environmental mapping using only the "background noise" of standard data transmissions, without adding expensive new hardware?

Methods

The authors propose a pipeline that treats a Large Intelligent Surface (LIS) as a massive electromagnetic aperture.

  1. Radio Map Generation: They model the LIS as a massive grid of antenna elements (259x259). Because the surface is so large, they use a spherical-wave Matched Filter (MF). Unlike traditional far-field assumptions (plane waves), this accounts for the curvature of radio waves in indoor settings, allowing for much sharper focus.
  2. Image Processing Pipeline: Once the radio signal is converted into a 2D "radio map" (which looks like a heat map), they apply several computer vision techniques:
  3. K-means Clustering: To binarize the image, separating high-energy reflections from background noise.
  4. Template Matching: To identify and "subtract" the known signal pattern of the active transmitter, leaving only the reflections from the environment.
  5. Masking & Logical OR: By comparing current maps to a pre-scanned "empty room" map, they can filter out static objects (like furniture) to isolate moving humans.
  6. Component Labeling: To count and locate individual "blobs" representing different people.

Datasets

The study relies on Ray Tracing simulations performed using the Feko software suite.

  • Scenario: A 10.34m x 10.34m room.
  • Targets: Passive humans modeled as rectangles with realistic dielectric properties, and active users modeled as narrowband transmitters at 3.5 GHz.
  • Ground Truth: The simulation provides exact coordinates for all objects, allowing for precise calculation of positioning errors.

Key Insights and Ideas

  • The LIS as a Lens: The most significant insight is that a sufficiently large antenna array acts exactly like a camera lens. The "aperture" of the LIS allows it to resolve spatial details that smaller, traditional antennas simply cannot see.
  • Passive Sensing via "Virtual Sources": Humans don't need to carry a phone to be detected. They act as "virtual sources" by reflecting ambient signals. The LIS is sensitive enough to pick up these weak reflections and treat them as distinct objects.
  • Robustness to Noise: Because the Matched Filter maximizes the Signal-to-Noise Ratio (SNR), the system performs remarkably well even when the environment is "noisy" (0 dB SNR), meaning the background interference is as strong as the signal itself.

Key Results

  • Active User Detection: The system detected 100% of active users with a median positioning error of only 5.5 cm.
  • Passive Human Detection: In a scenario with 10 passive humans, the system detected 98% of them.
  • Positioning Accuracy: The average positioning error for passive humans was approximately 28 cm, which is highly impressive for a system not using dedicated radar pulses.
  • High Resolution: The system could distinguish between two people standing as close as 25 cm apart, though detection reliability was highest when people were separated by 1 meter or more.

Limitations and Future Work

  • Static Background Requirement: The current method requires an initial "offline" scan of the empty room to create a masking map. In a real-world dynamic environment (like a busy mall), creating this baseline is difficult.
  • Computational Complexity: Processing a 259x259 antenna array (over 67,000 elements) in real-time requires massive computational power that isn't yet standard in base stations.
  • Narrowband Focus: The study uses narrowband signals. Future work could explore how wideband signals (which provide better time-of-flight data) could further improve depth perception and resolution.
  • Location-Assisted Beamforming: The authors suggest using these radio maps to proactively "aim" communication beams at users, potentially solving the beam-alignment challenge in 6G.

Summary updated

Abstract

This paper leverages the potential of Large Intelligent Surface (LIS) for radio sensing in 6G wireless networks. Major research has been undergone about its communication capabilities but it can be exploited as a formidable tool for radio sensing. By taking advantage of arbitrary communication signals occurring in the scenario, we apply direct processing to the output signal from the LIS to obtain a radio map that describes the physical presence of passive devices (scatterers, humans) which act as virtual sources due to the communication signal reflections. We then assess the usage of machine learning (k-means clustering), image processing and computer vision (template matching and component labeling) to extract meaningful information from these radio maps. As an exemplary use case, we evaluate this method for both active and passive user detection in an indoor setting. The results show that the presented method has high application potential as we are able to detect around 98% of humans passively and 100% active users by just using communication signals of commodity devices even in quite unfavorable Signal-to-Noise Ratio (SNR) conditions.
BibTeX
@inproceedings{vacarubio2021radio,
  author = {Cristian J. Vaca-Rubio and Pablo Ramirez-Espinosa and Kimmo Kansanen and Zheng‐Hua Tan and Elisabeth de Carvalho},
  title = {{Radio Sensing with Large Intelligent Surface for 6G}},
  booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing},
  year = {2021},
}
Figure