Device-Free 3D Drone Localization in RIS-Assisted mmWave MIMO Networks

Basic Information

• Paper ID: 2404.14879
• Title: Device-Free 3D Drone Localization in RIS-Assisted mmWave MIMO Networks
• Authors: Jiguang He, Charles Vanwynsberghe, Hui Chen, Chongwen Huang, Aymen Fakhreddine
• Category: eess.SP (Signal Processing)
• Publication Date: April 2024 (arXiv preprint, updated January 2, 2025)
• Paper Link: https://arxiv.org/abs/2404.14879

Abstract

This paper investigates the potential of Reconfigurable Intelligent Surfaces (RIS) for enabling passive/device-free three-dimensional (3D) drone localization in millimeter-wave (mmWave) frequency bands using multi-antenna cellular infrastructure. The developed localization system operates in bistatic mode without requiring direct communication between the drone and the base station. Theoretical performance limits are analyzed through Fisher information analysis and Cramér-Rao Lower Bound (CRLB) analysis. Furthermore, a low-complexity yet effective drone localization algorithm is developed based on coordinate gradient descent, and the impacts of drone radar cross-section (RCS) and training overhead on system performance are investigated. Both theoretical analysis and numerical simulations demonstrate significant advantages of RIS-integrated systems compared to RIS-free systems.

Research Background and Motivation

Problem Definition

The core problem addressed in this paper is: achieving passive three-dimensional localization of airborne drones using existing cellular infrastructure, i.e., accurately determining the spatial position of a drone without active communication between the drone and the base station.

Problem Significance

1. Security Requirements: With the proliferation of drone applications, the need for detection and localization of unauthorized or malicious drones is increasingly urgent
2. Infrastructure Reuse: Fully leveraging existing 5G/6G cellular network infrastructure while avoiding deployment of dedicated radar systems
3. Technology Integration: An important application scenario for the Integrated Sensing and Communication (ISAC) paradigm

Limitations of Existing Methods

1. Traditional Radar Systems: Require specialized radar equipment with high costs and limited coverage
2. Active Localization Systems: Require communication between targets and base stations, unsuitable for non-cooperative targets
3. Existing Passive Systems: Limited performance in mmWave bands with insufficient localization accuracy

Research Motivation

1. Leveraging RIS to enhance signal propagation paths and improve passive localization performance
2. Achieving high-precision 3D localization in mmWave MIMO systems
3. Developing low-complexity practical algorithms approaching theoretical performance limits

Core Contributions

1. System Design: Proposes an RIS-assisted mmWave MIMO passive drone localization system architecture
2. Theoretical Analysis: Analyzes theoretical performance limits through Fisher information matrix and CRLB analysis
3. Algorithm Development: Develops a low-complexity three-stage localization algorithm based on coordinate gradient descent (CGD)
4. Performance Evaluation: Comprehensively analyzes the impact of key parameters such as RCS and training overhead on system performance
5. Advantage Verification: Demonstrates through theory and simulation the significant advantages of RIS-assisted systems compared to conventional methods

Methodology Details

Task Definition

Input: Multi-slot received signals at the receiving user equipment (UE) Output: Three-dimensional coordinates of the drone $(x_D, y_D, z_D)$Constraints: No direct communication between drone and base station; system operates in bistatic mode

System Architecture

Network Topology

The system comprises three key components:

• Millimeter-Wave Base Station (BS): Equipped with $M_B = M_{B,y} \times M_{B,z}$ antenna uniform planar array
• Reconfigurable Intelligent Surface (RIS): Contains $M_R = M_{R,x} \times M_{R,y}$ reflecting elements
• Receiving User Equipment (UE): Equipped with $M_U$ antennas

Channel Modeling

Employs the Saleh-Valenzuela channel model considering sparse scattering characteristics in mmWave bands. Key channels include:

1. BS-RIS Channel $\mathbf{H}_1 \in \mathbb{C}^{M_R \times M_B}$: $\mathbf{H}_1 = \frac{e^{-j2\pi d_1/\lambda}}{\sqrt{\rho_1}} \boldsymbol{\alpha}_x(\theta_{r,1}, \phi_{r,1}) \otimes \boldsymbol{\alpha}_y(\theta_{r,1}, \phi_{r,1}) \left(\boldsymbol{\alpha}_y(\theta_{t,1}, \phi_{t,1}) \otimes \boldsymbol{\alpha}_z(\phi_{t,1})\right)^H$
2. BS-Drone Channel $\mathbf{h}_2$, RIS-Drone Channel $\mathbf{h}_3$, Drone-UE Channel $\mathbf{h}_4$, etc.

Signal Transmission Model

Downlink Probing Process

The base station employs multi-beam methods:

• Fixed beam $\mathbf{f}_0$ directed toward RIS
• Time-varying beam $\mathbf{f}_k$ directed toward sky region
• RIS time-varying beam $\boldsymbol{\omega}_k$ covering target region

Received signal model: $\mathbf{y}_k = \zeta\mathbf{h}_4(\mathbf{h}_3\text{diag}(\boldsymbol{\omega}_k)\mathbf{H}_1\mathbf{F}_k\mathbf{s}_k + \mathbf{h}_2\mathbf{F}_k\mathbf{s}_k) + \mathbf{H}_5\mathbf{F}_k\mathbf{s}_k + \mathbf{n}_k$

where $\zeta$ is the radar cross-section (RCS) of the drone.

Three-Stage Localization Algorithm

Stage One: Channel Estimation

Employs coordinate gradient descent (CGD) method to estimate cascaded channels:

Objective function: $\min f(\mathbf{h}_2,\mathbf{h}_3,\mathbf{h}_4) = \left\|\mathbf{Y} - \sqrt{P/2}\tilde{\mathbf{H}}\bar{\boldsymbol{\Omega}} - \sqrt{P/2}\hat{\mathbf{H}}\bar{\mathbf{F}}\right\|_F^2$

Iterative update: $\mathbf{h}_4^{(k)} = \mathbf{h}_4^{(k-1)} - \eta \frac{\partial f}{\partial \mathbf{h}_4}\bigg|_{\mathbf{h}_4=\mathbf{h}_4^{(k-1)}}$

Stage Two: Angle Parameter Estimation

Extracts angle parameters through 2D search: $\hat{\theta}_{t,2}, \hat{\phi}_{t,2} = \arg\max_{\theta_{t,2},\phi_{t,2}} \left|\hat{\mathbf{h}}_2(\boldsymbol{\alpha}_y(\theta_{t,2}, \phi_{t,2}) \otimes \boldsymbol{\alpha}_z(\phi_{t,2}))\right|$

Stage Three: Position Mapping

Maps angle estimates to 3D position based on least squares principle: $\hat{\mathbf{p}}_D = (\mathbf{B}_B + \mathbf{B}_R + \mathbf{B}_U)^{-1}(\mathbf{B}_B\mathbf{p}_B + \mathbf{B}_R\mathbf{p}_R + \mathbf{B}_U\mathbf{p}_U)$

Technical Innovations

1. Multi-Path Utilization: Simultaneously exploits both BS-Drone-UE and BS-RIS-Drone-UE paths
2. Beam Design: Ensures base station beam orthogonality to reduce direct path interference
3. Low-Complexity Algorithm: CGD method offers lower computational complexity compared to nuclear norm minimization
4. Theoretical Bounds: Provides Fisher information analysis and CRLB as performance benchmarks

Experimental Setup

Simulation Parameters

• Antenna Configuration: $M_B = 8 \times 8$, $M_U = 4 \times 4$, $M_R = 6 \times 6$
• Node Positions:
  • BS: (0, 0, 26)
  • RIS: (0, 0.5, 25.5)
  • Drone: (3, 3, 30)
  • UE: (2, 2, 24)
• Channel Parameters: Path loss exponent $\Gamma = 2$, bandwidth $B = 20$ MHz
• Noise Variance: $\sigma^2 = -101$ dBm

Evaluation Metrics

Root Mean Square Error (RMSE) is employed as the primary evaluation metric: $\text{RMSE} = \sqrt{E\{(\mathbf{p}_D - \hat{\mathbf{p}}_D)^T(\mathbf{p}_D - \hat{\mathbf{p}}_D)\}}$

Comparison Methods

• With RIS vs. Without RIS: Validates RIS performance gains
• Theoretical CRLB: Serves as performance upper bound benchmark

Experimental Results

Main Results

RIS Performance Gains

Simulation results in the SNR range of -10 dB to 10 dB show:

• RIS-Equipped System: Demonstrates significant advantages in low SNR regions
• Performance Gap: At -10 dB SNR, RIS system achieves approximately one order of magnitude RMSE reduction compared to RIS-free systems
• Theoretical Proximity: The proposed algorithm performance approaches the theoretical CRLB lower bound

Training Overhead Impact

Impact of different numbers of training slots K:

• Larger K values yield higher localization accuracy
• Performance saturation point exists; marginal gains from excessive K increase diminish

RCS Impact Analysis

Significant impact of radar cross-section on performance:

• $\zeta = 0.5$: Relatively poor performance
• $\zeta = 1.0$: Standard performance baseline
• $\zeta = 2.0$: Significant performance improvement

Key Findings

1. Low SNR Advantage: RIS demonstrates most significant performance gains under low signal-to-noise ratio conditions
2. RCS Sensitivity: Drone radar cross-section is a critical factor affecting localization accuracy
3. Algorithm Effectiveness: CGD algorithm achieves near-theoretical-limit performance with low complexity

Related Work

Main Research Directions

1. 5G/6G Localization Technologies: Localization methods based on mmWave and large-scale MIMO
2. Radar-Communication Fusion: Target detection and localization in ISAC systems
3. RIS-Assisted Communications: Applications of reconfigurable intelligent surfaces in wireless communications

Positioning of This Work's Contributions

• Compared to traditional radar: Utilizes existing cellular infrastructure, reducing deployment costs
• Compared to active localization: Applicable to non-cooperative targets, expanding application scope
• Compared to existing RIS research: First application of RIS to passive drone localization scenarios

Conclusions and Discussion

Main Conclusions

1. RIS significantly enhances passive drone localization performance in mmWave systems
2. The proposed CGD algorithm achieves excellent performance with low complexity
3. System parameters (RCS, training overhead) have important impacts on performance

Limitations

1. Channel Assumptions: Only considers line-of-sight (LoS) paths; actual environments may have multipath propagation
2. Static Scenarios: Does not consider impact of drone movement on localization performance
3. Perfect Synchronization: Assumes perfect time and frequency synchronization
4. Known Positions: Assumes RIS and UE positions are precisely known

Future Directions

1. Power Allocation Optimization: Research optimal power allocation strategies among base station multi-beams
2. Dynamic Tracking: Extension to trajectory tracking of mobile drones
3. Multipath Processing: Robust localization considering non-line-of-sight propagation environments
4. Practical Deployment: Validation of algorithm performance in real-world environments

In-Depth Evaluation

Strengths

1. Strong Innovation: First application of RIS technology to passive drone localization with significant novelty
2. Complete Theory: Provides comprehensive Fisher information analysis and CRLB derivation
3. Practical Algorithm: CGD algorithm has low complexity and is amenable to practical implementation
4. Comprehensive Analysis: Systematically analyzes impact of key parameters on performance
5. Sufficient Verification: Theoretical analysis and simulation results mutually validate each other

Weaknesses

1. Idealized Assumptions: LoS channel assumptions may not hold in practical environments
2. Limited Scenarios: Primarily applicable to high-altitude drone localization in open spaces
3. Experimental Limitations: Lacks real-world testing validation; only simulation results provided
4. Interference Handling: Insufficient consideration of interference suppression in complex electromagnetic environments

Impact and Significance

1. Academic Value: Provides new perspectives for RIS applications in sensing research
2. Practical Prospects: Potential applications in smart cities, security monitoring, and other fields
3. Technology Advancement: Promotes development and standardization of ISAC technology

Applicable Scenarios

• Drone monitoring systems on urban high-rise rooftops
• Drone detection networks around airports
• Airspace security protection for critical infrastructure
• Value-added service applications in 5G/6G networks

References

The paper cites multiple important works in related fields, including 5G localization technologies, ISAC systems, and RIS communications, providing a solid theoretical foundation for this work.