Integrated Sensing and Communication With Massive MIMO: A Unified Tensor Approach for Channel and Target Parameter Estimation

Ruoyu Zhang, Lei Cheng, Shuai Wang, Yi Lou, Yulong Gao, Wen Wu, Derrick Wing Kwan Ng

TL;DR

This paper introduces a unified mathematical framework that treats communication channel estimation and radar target sensing as the same problem. By modeling received signals as multi-dimensional tensors and applying Canonical Polyadic Decomposition (CPD), the authors achieve high-accuracy parameter estimation (angle, delay, Doppler) for both tasks simultaneously, even in wideband systems plagued by "beam squint" effects.

What Problem Does This Paper Address?

In the push toward 6G, Integrated Sensing and Communication (ISAC) aims to let wireless networks do double duty: transmitting data to users while simultaneously acting as a radar to "sense" the environment.

However, massive MIMO systems (which use huge antenna arrays) face a "dimensionality curse." Acquiring accurate Channel State Information (CSI) for communication and Target Parameter Information (TPI) for sensing usually requires massive training overhead, eating up the very spectral efficiency ISAC is supposed to save. Furthermore, in wideband systems, a phenomenon called beam squint occurs—where the signal's spatial direction appears to shift across different frequencies—making traditional estimation methods fail.

Methods

The authors propose a Unified Tensor Approach. Instead of treating the communication channel as a giant, abstract matrix, they parameterize it using physical attributes: angles of arrival/departure, time delays, and Doppler shifts. This makes the communication model look mathematically identical to a radar sensing model.

Key Technical Innovations:

  1. Tensor Modeling: They formulate the received signals as a third-order tensor (Space $\times$ Time $\times$ Frequency). This structure is then decomposed using Canonical Polyadic Decomposition (CPD).
  2. Exploiting Vandermonde Structure: The authors leverage the mathematical "Vandermonde" structure inherent in uniform linear arrays. This allows them to prove a "relaxed uniqueness condition," meaning they can resolve more targets/paths than standard tensor methods.
  3. Iterative Refinement: They developed a unified algorithm that iteratively decouples the estimation of the Angle of Departure (AoD) and Doppler shift, which are often "tangled" in the signal.
  4. Segment-Based Training: To combat beam squint, they designed a training pattern that repeats precoders over segments. This "freezes" the frequency-dependent steering vectors within a tensor framework, allowing them to decouple the wideband effects.

Datasets

The paper relies on numerical simulations rather than a public real-world dataset (which is standard for this sub-field of information theory).

  • Setup: A Base Station with 64 transmit antennas and 8 receive antennas, serving a User Equipment (UE) with 8 antennas.
  • Parameters: Carrier frequency at 28 GHz (mmWave), bandwidth up to 1 GHz (to test beam squint), and targets moving at speeds up to 120 km/h.
  • Benchmarks: The proposed algorithms were compared against classic methods like MUSIC (Multiple Signal Classification), Matched Filtering (MF), and standard Alternating Least Squares (ALS) tensor methods.

Key Insights and Ideas

  • The Physics-Comms Link: The "aha moment" is that a multipath communication channel is essentially just a collection of "targets" (reflectors) that happen to be cooperative. By parameterizing the channel physically, you can use radar tools for comms and vice versa.
  • Tensors > Matrices: Traditional matrix-based estimation flattens the signal, losing the inherent correlations across space, time, and frequency. Tensors preserve this multi-dimensional "geometry," leading to much higher resolution.
  • Breaking the Resolution Limit: By exploiting the Vandermonde structure, the authors showed they could resolve up to 20 targets in a setup where traditional MUSIC-based methods failed after just 11.

Key Results

  • Superior Resolution: The proposed method maintains a near 100% success rate in distinguishing targets even as the number of targets increases, whereas the success rate of the MUSIC algorithm drops to zero once the number of targets exceeds the number of receive antennas.
  • High Accuracy: The Root Mean Square Error (RMSE) for angle, range, and velocity estimation approaches the Cramer-Rao Bound (CRB)—the theoretical physical limit of how accurate an estimate can be.
  • Reduced Overhead: The unified approach achieves better performance with significantly fewer training subcarriers ($K=16$) compared to benchmark algorithms that required $K=32$.
  • Robustness to Beam Squint: In 1 GHz wideband scenarios, the segment-based algorithm prevented the "performance floor" (where error stops decreasing despite higher SNR) that hit all other benchmark methods.

Limitations and Future Work

  • Interference Classification: While the model handles interference, it currently struggles to distinguish between a "real" target of interest and "clutter" (non-target objects like trees or buildings) without prior knowledge. The authors suggest target classification as a future research path.
  • Computational Complexity: The iterative refinement for AoD and Doppler estimation is more computationally intensive than simple matched filtering. While they provide a lower-complexity factor matrix estimation, the final parameter extraction still requires significant cycles.
  • Hardware Non-Idealities: The study assumes relatively "clean" hardware. Future work could explore how phase noise and power amplifier non-linearities affect the tensor decomposition's stability.

Summary updated

Abstract

Benefitting from the vast spatial degrees of freedom, the amalgamation of integrated sensing and communication (ISAC) and massive multiple-input multiple-output (MIMO) is expected to simultaneously improve spectral and energy efficiencies as well as the sensing capability. However, a large number of antennas deployed in massive MIMO-ISAC raises critical challenges in acquiring both accurate channel state information and target parameter information. To overcome these two challenges with a unified framework, we first analyze their underlying system models and then propose a novel tensor-based approach that addresses both the channel estimation and target sensing problems. Specifically, by parameterizing the high-dimensional communication channel exploiting a small number of physical parameters, we associate the channel state information with the sensing parameters of targets in terms of angular, delay, and Doppler dimensions. Then, we propose a shared training pattern adopting the same time-frequency resources such that both the channel estimation and target parameter estimation can be formulated as a canonical polyadic decomposition problem with a similar mathematical expression. On this basis, we first investigate the uniqueness condition of the tensor factorization and the maximum number of resolvable targets by utilizing the specific Vandermonde structure. Then, we develop a unified tensor-based algorithm to estimate the parameters including angles, time delays, Doppler shifts, and reflection/path coefficients of the targets/channels. In addition, we propose a segment-based shared training pattern to facilitate the channel and target parameter estimation for the case with significant beam squint effects. Simulation results verify our theoretical analysis and the superiority of the proposed unified algorithms in terms of estimation accuracy, sensing resolution, and training overhead reduction.
BibTeX
@article{zhang2024integrated,
  author = {Ruoyu Zhang and Lei Cheng and Shuai Wang and Yi Lou and Yulong Gao and Wen Wu and Derrick Wing Kwan Ng},
  title = {{Integrated Sensing and Communication With Massive MIMO: A Unified Tensor Approach for Channel and Target Parameter Estimation}},
  journal = {IEEE Trans. Wireless Communications},
  year = {2024},
  url = {https://openreview.net/forum?id=T8z3YBMOzX},
}

Loading related content...

Figure