Uplink Sensing Using CSI Ratio in Perceptive Mobile Networks

Zhitong Ni, J. Andrew Zhang, Kai Wu, Ren Ping Liu

TL;DR

This paper introduces a novel method for "uplink sensing" in mobile networks—using the signals sent from phones to base stations to detect and track moving objects. By applying a multi-element Taylor series to the CSI (Channel State Information) ratio, the authors successfully cancel out hardware synchronization errors and localize multiple moving targets simultaneously, even in challenging environments without a clear line-of-sight.

What Problem Does This Paper Address?

In Perceptive Mobile Networks (PMNs), the goal is to turn the ubiquitous cellular infrastructure into a giant radar system. Uplink sensing is particularly attractive because it uses existing signals sent by users (UEs) to the base station (BS) to sense the environment.

However, two major "deal-breakers" usually get in the way:

  1. Clock Asynchronism: UEs and BSs are rarely perfectly synchronized. This creates Timing Offsets (TO) and Carrier Frequency Offsets (CFO) that act as noise, masking the subtle phase changes caused by moving targets.
  2. Line-of-Sight (LOS) Dependency: Most existing "CSI ratio" techniques (which divide signals from different antennas to cancel out offsets) require a strong, direct LOS path to work. In real-world urban or indoor environments, that LOS path is often blocked.

This paper aims to break these limitations, allowing for the extraction of Doppler frequency, Angle-of-Arrival (AoA), and delay for multiple targets in asynchronous, Non-Line-of-Sight (NLOS) scenarios.

Methods

The core innovation is the mathematical treatment of the CSI Ratio.

  • The Ratio Trick: By dividing the CSI received at one antenna by the CSI from another, common hardware offsets (TO and CFO) are mathematically canceled out.
  • Multi-element Taylor Series: The CSI ratio is a complex, nonlinear function of the target's parameters. The authors use a truncated Taylor series to linearize this function. This transformation is the "key" that unlocks the use of traditional, high-resolution sensing algorithms.
  • D-CSIR (Difference of CSI-Ratio): They introduce the concept of the "Difference of CSI-ratio" to isolate dynamic (moving) paths from the static background clutter.
  • Parameter Extraction:
    • Doppler & AoA: Once linearized, the authors apply MUSIC (Multiple Signal Classification), a popular subspace-based algorithm, to estimate the speeds and angles of multiple targets.
  • Delay: They use a Least-Squares (LS) approach to estimate the propagation delay, which is essential for determining the target's distance.

Datasets

The authors validated their approach using two distinct methods:

  1. Experimental Data: They used raw data collected from a 3-antenna Commercial Off-The-Shelf (COTS) WiFi device operating at 5.32 GHz. This involved tracking a single moving human in an indoor environment.
  2. Numerical Simulations: They used MATLAB to simulate a more complex environment (3 GHz carrier, 64 subcarriers) with one dynamic path and five static paths. This allowed them to test the algorithm's performance across various Signal-to-Noise Ratios (SNR) and multi-target scenarios.

Key Insights and Ideas

  • Spatial Division is Key: The paper highlights that offsets can only be fully removed by dividing signals across the spatial domain (antennas). Dividing across time or frequency domains actually re-introduces the offsets into the equations.
  • Linearization Enables Complexity: By turning a nonlinear sensing problem into a linear one via the Taylor series, the authors moved beyond simple single-target Doppler estimation to full multi-parameter, multi-target localization.
  • Static Paths are a Feature, Not a Bug: Interestingly, the Taylor series converges better when the static paths (reflections from walls, etc.) are stronger than the dynamic paths. Since static reflections are usually dominant in wireless channels, this makes the method highly robust for real-world use.
  • NLOS Capability: Unlike previous CSI-ratio methods, this scheme does not require a dominant LOS path, making it viable for "around-the-corner" sensing.

Key Results

  • Trajectory Accuracy: In real-world WiFi tests, the proposed method tracked a human target with high precision, closely matching "ground truth" data obtained from a high-end mmWave radar.
  • Superiority over Benchmarks: In simulations, the proposed estimator significantly outperformed existing methods like AMS and CACC-MUSIC, particularly in NLOS scenarios where the benchmarks' performance collapsed.
  • Resolution: The vectorized manifold approach used for AoA estimation increased the angular resolution, allowing the system to distinguish between targets even with a limited number of antennas.

Limitations and Future Work

  • Delay Sensitivity: Estimating the absolute delay (distance) remains the most challenging part. It requires either knowing the static component of the channel or observing the channel over a long training period to "average out" the static parts.
  • Computational Load: While the Taylor series simplifies the math, grid-searching in MUSIC-type estimators can be computationally expensive as the number of targets and parameters increases.
  • Future Directions: The authors suggest that future work could involve optimizing the algorithm for even more complex environments with higher numbers of UEs and more cluttered dynamic scenes.

Summary updated

Abstract

Uplink sensing in perceptive mobile networks (PMNs), which uses uplink communication signals for sensing the environment around a base station, faces challenging issues of clock asynchronism and the requirement of a line-of-sight (LOS) path between transmitters and receivers. The channel state information (CSI) ratio has been applied to resolve these issues, however, current research on the CSI ratio is limited to Doppler estimation in a single dynamic path. This paper proposes an advanced parameter estimation scheme that can extract multiple dynamic parameters, including Doppler frequency, angle-of-arrival (AoA), and delay, in a communication uplink channel and completes the localization of multiple moving targets. Our scheme is based on the multi-element complex Taylor series of the CSI ratio that converts a nonlinear function of sensing parameters to linear forms and enables the applications of traditional sensing algorithms. Using the truncated Taylor series, we develop novel multiple-signal-classification grid searching algorithms for estimating Doppler frequencies and AoAs and use the least-square method to obtain delays. Both experimental and simulation results are provided, demonstrating that our proposed scheme can achieve good performances for sensing both single and multiple dynamic paths, without requiring the presence of a LOS path.
BibTeX
@article{ni2023uplink,
  author = {Zhitong Ni and J. Andrew Zhang and Kai Wu and Ren Ping Liu},
  title = {{Uplink Sensing Using CSI Ratio in Perceptive Mobile Networks}},
  journal = {IEEE Transactions on Signal Processing},
  year = {2023},
  doi = {10.1109/tsp.2023.3294626},
}
Figure