Revolutionizing RIS Networks: LiDAR-Based Data-Driven Approach to Enhance RIS Beamforming

Area of Science:

• Wireless Communication
• Signal Processing
• Artificial Intelligence

Background:

• Reconfigurable Intelligent Surfaces (RIS) are crucial for next-generation networks.
• Accurate user localization and beamforming are key challenges in RIS-assisted networks.
• Existing methods struggle with precise positioning and optimal RIS configuration.

Purpose of the Study:

• To propose a data-driven approach for enhanced user localization and beamforming in RIS-assisted networks.
• To leverage Light Detecting and Ranging (LiDAR) sensors for precise 3D mapping and user positioning.
• To optimize RIS reflection coefficients for maximizing the sum rate in multi-user scenarios.

Main Methods:

• Integration of LiDAR sensors for high-speed, precise 3D mapping and user localization.
• Extension of Graph Neural Networks (GNNs) by incorporating LiDAR-captured user locations.
• Training the GNN to learn the mapping from received pilots to optimal beamformers and RIS coefficients.
• Utilizing the permutation-equivariant and -invariant properties of GNNs for efficient LiDAR data handling.

Main Results:

• Significant improvements in sum rates achieved compared to conventional methods.
• Incorporating user locations enhanced performance by up to 25% compared to excluding them.
• Outperformed Linear Minimum Mean Squared Error (LMMSE) channel estimation by up to 98% with varying pilot lengths.
• Demonstrated a 190% increase in sum rate with varying downlink power compared to scenarios without RIS.

Conclusions:

• The proposed LiDAR-enhanced GNN approach effectively optimizes RIS-assisted networks.
• Precise user localization via LiDAR is instrumental in maximizing network performance.
• This method offers substantial gains in sum rate and outperforms existing channel estimation techniques.