Variational Bayesian Multiuser Tracking for Reconfigurable Intelligent Surface-Aided MIMO-OFDM Systems

Xiaojun Yuan, Boyu Teng, Xiaojun Yuan 0002, Rui WangIEEE JSAC 2023 · DOI

TL;DR

This paper tackles the challenging problem of precisely tracking multiple users in a complex wireless environment using Reconfigurable Intelligent Surfaces (RIS) and MIMO-OFDM systems. They introduce a novel Bayesian tracking algorithm that accounts for potential line-of-sight blockages and optimizes RIS beamforming for improved accuracy, achieving centimeter-level tracking.

What Problem Does This Paper Address?

Modern wireless systems, especially those aiming for 6G, are moving towards integrated sensing and communication (ISAC). A critical aspect of sensing is high-precision localization and tracking of users. Traditional methods often struggle in complex environments with signal blockages. Reconfigurable Intelligent Surfaces (RIS) offer a promising solution by creating controllable reflective paths to enhance signal reception and localization. However, accurately tracking multiple users simultaneously, especially when line-of-sight (LoS) paths can be blocked by obstacles, remains a significant challenge. This paper addresses this by developing a robust tracking system that leverages multiple RISs in a MIMO-OFDM setup, explicitly modeling and handling LoS path blockages.

Methods

The core of this work is the Bayesian Multiuser Tracking (BMT) algorithm. This algorithm is built upon a probabilistic model that captures user mobility and the dynamic nature of LoS path existence (birth and death processes due to blockages). Key methodological contributions include:

  • Probability Transition Model: A model is established to characterize user movement and the intermittent nature of LoS paths between users and RISs.
  • Online Message Passing: A sum-product message passing algorithm is employed on a factor graph representation of the probabilistic model. This allows for iterative estimation of user positions, angles of arrival (AoAs) at RISs, time delays, and LoS path existence.
  • Variational Inference and Gaussian Approximation: To manage the computational complexity, variational inference and Gaussian approximation are used to calculate messages.
  • Bayesian Cramér Rao Bound (BCRB): The BCRB is derived to establish a theoretical lower bound on the tracking performance, serving as a benchmark.
  • BCRB-based Passive Beamforming (PBF) Design: The derived BCRB is used to optimize the PBF of the RISs, aiming to maximize tracking accuracy. This optimization problem is reformulated using estimations from the previous frame due to the unavailability of true parameters.

Datasets

Not discussed in the paper. The experiments are based on simulated data generated according to the system model and parameters defined in the paper.

Key Insights and Ideas

  • RISs as Dynamic Anchors: RISs are not just passive reflectors but can act as dynamic, configurable anchors that significantly improve localization accuracy, especially when LoS paths are unreliable.
  • Blockage Modeling is Crucial: Explicitly modeling the birth and death of LoS paths is essential for robust tracking in realistic mobile scenarios, preventing performance degradation when obstacles appear.
  • Joint Optimization of Tracking and Beamforming: Optimizing RIS beamforming based on the theoretical performance limits (BCRB) directly leads to substantial improvements in tracking accuracy.
  • Centimeter-Level Accuracy Achievable: The proposed BMT algorithm, combined with optimized PBF, demonstrates the potential for achieving very high tracking precision (centimeter-level).

Key Results

  • The proposed BMT algorithm achieves centimeter-level tracking accuracy.
  • The BCRB-based PBF design significantly outperforms random PBF and DFT codebook-based PBF schemes in terms of multiuser tracking performance.
  • Increasing the number of RIS elements (NR) leads to higher tracking performance due to improved SNR and angle domain resolution.
  • The performance of the BMT algorithm is sensitive to user mobility; stronger mobility leads to weaker temporal correlation and poorer performance, though it remains close to the BCRB.
  • The accuracy of LoS path blockage estimation is close to 1 across a reasonable SNR range.

Limitations and Future Work

  • Computational Complexity: While the paper proposes a low-complexity approximate solution, the variational inference approach still involves significant computation, especially with a large number of RIS elements or users.
  • Initialization: The algorithm's performance in the initial frame relies on accurate initial position information (e.g., from GPS).
  • Assumptions: The paper assumes a quasi-static flat fading channel model within a pilot frame and a specific geometric channel model.
  • Future Work: The paper suggests exploring more advanced optimization techniques for PBF and potentially extending the framework to more complex channel models or scenarios. The authors also highlight the potential for further improvements by considering more sophisticated RIS configurations or integration with other sensing modalities.

Summary updated

BibTeX
@article{yuan2023variational,
  author = {Xiaojun Yuan and Boyu Teng and Xiaojun Yuan and Rui Wang},
  title = {{Variational Bayesian Multiuser Tracking for Reconfigurable Intelligent Surface-Aided MIMO-OFDM Systems}},
  journal = {IEEE JSAC},
  year = {2023},
  doi = {10.1109/jsac.2023.3322792},
}
Figure