Joint Beamforming for RIS-Assisted Integrated Sensing and Communication Systems

Yongqing Xu, Yong Li, J. Andrew Zhang, Marco Di Renzo, Tony Q. S. Quek

TL;DR

This paper introduces two novel optimization frameworks to manage the "tug-of-war" between sensing and communication in 6G networks using Reconfigurable Intelligent Surfaces (RIS). By adding an RIS to an Integrated Sensing and Communication (ISAC) system, the authors provide the extra degrees of freedom needed to maximize radar sensing accuracy without sacrificing high-speed data transmission.

What Problem Does This Paper Address?

Future wireless networks (6G) are expected to do more than just move data; they need to "see" the world using radar-like sensing for applications like autonomous driving and smart factories. This is known as Integrated Sensing and Communication (ISAC).

The fundamental challenge is a resource conflict: sensing and communication often require different beamforming "shapes" to perform optimally. In a standard system, you only have a limited number of antennas (degrees of freedom) to satisfy both. This paper tackles this bottleneck by introducing a Reconfigurable Intelligent Surface (RIS)—a "smart mirror" that can shift the phase of incoming signals. The RIS provides additional paths for both the communication signal and the radar echo, effectively expanding the optimization space to improve both metrics simultaneously.

Methods

The authors propose two distinct Alternating Optimization (AO) algorithms to solve the joint beamforming problem (coordinating the Base Station's antennas and the RIS's reflecting elements):

  1. AO-SDR-ODI: This method focuses on maximizing Radar Mutual Information (MI) while maintaining a strict minimum requirement for communication rates. It uses Semidefinite Relaxation (SDR) for the transmit beamforming and a novel One-Dimension Iterative (ODI) algorithm to find the best RIS phase shifts. While highly accurate, its complexity grows quickly with the number of RIS elements.
  2. AO-RG (Riemannian Gradient): To handle larger-scale systems, the authors developed a manifold-based approach. By treating the hardware constraints (like the constant-modulus constraint of RIS elements) as a geometric surface (a manifold), they can use gradient-based optimization. This method maximizes a weighted sum of both sensing and communication performance and is much more computationally efficient for large surfaces.

Datasets

As is standard for theoretical physical-layer research, the paper relies on numerical simulations rather than a static dataset. The authors modeled a Multi-Input Multi-Output (MIMO) environment with:

  • A dual-functional Base Station (4 transmit/receive antennas).
  • Multiple single-antenna users.
  • Multiple scattering targets for radar sensing.
  • An RIS with varying numbers of elements ($M$ ranging from 16 to 128).
  • Channel models accounting for both Line-of-Sight (LoS) and Rician fading.

Key Insights and Ideas

  • Radar MI as a Proxy for Accuracy: The paper reinforces that maximizing Radar Mutual Information is a powerful design goal because it is inversely correlated with the Mean-Squared Error (MSE) of target estimation. If you maximize MI, you inherently make your "radar" more precise.
  • The RIS "Virtual Path" Advantage: The RIS doesn't just help communication; it creates a "cascaded" radar path (Base Station $\rightarrow$ RIS $\rightarrow$ Target $\rightarrow$ RIS $\rightarrow$ Base Station). This extra echo significantly boosts the signal-to-noise ratio for sensing.
  • Scalability vs. Optimality: The SDR-based method is the "gold standard" for performance but hits a computational wall. The Riemannian Gradient method offers a "fast-track" alternative that is nearly as good but scales much better for the massive RIS arrays expected in 6G.
  • Robustness to Quantization: The authors found that even with low-resolution RIS hardware (e.g., only 2 bits for phase control), the system still achieves most of the gains of an "ideal" continuous-phase RIS.

Key Results

  • Performance Gain: The proposed "Proposed MI" method significantly outperformed existing optimized MI benchmarks by better leveraging the RIS-reflected paths.
  • Convergence: The AO-RG algorithm proved highly efficient, typically converging in about 5 iterations, compared to 35 iterations for the more complex AO-SDR-ODI method.
  • Sensing Accuracy: Simulations confirmed that as Radar MI increases, the MSE of sensing parameters (like distance and angle) decreases sharply, validating the information-theoretic approach.
  • Power Efficiency: The system maintained high radar MI even when communication rate thresholds were pushed higher, proving that the RIS successfully mitigates the trade-off between the two functions.

Limitations and Future Work

  • Perfect CSI Assumption: The study assumes perfect Channel State Information (CSI). In real-world RIS deployments, estimating the channels between the BS, RIS, and users is notoriously difficult and overhead-heavy.
  • Static Targets: The current model focuses on stationary scatterers. Future work would need to account for Doppler shifts from moving targets (e.g., vehicles or drones).
  • Hardware Impairments: While the paper looks at discrete phase shifts, it does not fully account for other hardware non-idealities like phase-dependent amplitude variations in the RIS elements.
  • Interference Management: The authors suggest that more complex multi-user interference scenarios and robust design under imperfect CSI are the next logical steps for this research.

Summary updated

Abstract

Integrated sensing and communications (ISAC) is an emerging critical technique for the next generation of communication systems. However, due to multiple performance metrics used for communication and sensing, the limited degrees-of-freedom (DoF) in optimizing ISAC systems poses a challenge. Reconfigurable intelligent surfaces (RIS) can introduce new DoF for beamforming in ISAC systems, thereby enhancing the performance of communication and sensing simultaneously. In this paper, we propose two optimization techniques for beamforming in RIS-assisted ISAC systems. The first technique is an alternating optimization (AO) algorithm based on the semidefinite relaxation (SDR) method and a one-dimension iterative (ODI) algorithm, which can maximize the radar mutual information (MI) while imposing constraints on the communication rates. The second technique is an AO algorithm based on the Riemannian gradient (RG) method, which can maximize the weighted ISAC performance metrics. Simulation results verify the effectiveness of the proposed schemes. The AO-SDR-ODI method is shown to achieve better communication and sensing performance, than the AO-RG method, at a higher complexity. It is also shown that the mean-squared-error (MSE) of the estimates of the sensing parameters decreases as the radar MI increases.
BibTeX
@article{xu2024joint,
  author = {Yongqing Xu and Yong Li and J. Andrew Zhang and Marco Di Renzo and Tony Q. S. Quek},
  title = {{Joint Beamforming for RIS-Assisted Integrated Sensing and Communication Systems}},
  journal = {IEEE Trans. Communications},
  year = {2024},
  doi = {10.1109/tcomm.2023.3344143},
}
Figure