Efficient STAR-RIS Mode for Energy Minimization in WPT-FL Networks With NOMA

Mohammadhossein Alishahi, Ming Zeng, Paul Fortier, Omer Waqar, Muhammad Hanif, Dinh Thai Hoang, Diep N. Nguyen, Quoc-Viet PhamIEEE Trans. Communications 2025 · DOI

This paper, recently published by Alishahi et al., tackles one of the most pressing hurdles for the upcoming 6G era: how to keep billions of power-hungry Internet of Things (IoT) devices running without constantly swapping batteries or clogging the airwaves.

TL;DR

The researchers propose a novel framework that combines Simultaneously Transmitting and Reflecting Reconfigurable Intelligent Surfaces (STAR-RIS) with Wireless Power Transfer (WPT) and Federated Learning (FL). By jointly optimizing transmission modes and resource allocation, they demonstrate that a "Time Switching" approach for STAR-RIS can significantly minimize the total energy consumption of IoT networks compared to traditional methods.

What Problem Does This Paper Address?

As we move toward 6G, we are deploying massive numbers of IoT devices. These devices face a "triple threat":

  1. Energy Constraints: Small sensors have limited battery life.
  2. Communication Overhead: Sending raw data to the cloud for AI training is slow and energy-intensive.
  3. Coverage Gaps: Physical obstacles often block signals between devices and the central Access Point (AP).

To solve these, the authors integrate several cutting-edge technologies. Federated Learning allows devices to train AI models locally (keeping data private and reducing transmission), while Wireless Power Transfer beams energy to the devices to keep them alive. To fix coverage, they use STAR-RIS—a smart surface that doesn't just reflect signals (like conventional RIS) but can also transmit them through the surface, providing 360-degree coverage. The challenge lies in the math: how do you coordinate all these moving parts—power, time, frequency, and the smart surface's "phase shifts"—to use the absolute minimum amount of energy?

Methods

The authors formulate a complex, non-convex energy minimization problem. Because the variables (like beamforming and time allocation) are "coupled" (changing one affects the others), they use a decoupling strategy involving two main blocks:

  • Block A (Beamforming & Phase Shifts): This block optimizes the STAR-RIS phase shift vectors and the AP's beamforming matrices. They employ Block Coordinate Descent (BCD), alternating between solving Semi-Definite Programming (SDP) problems and Rayleigh Quotient (RQ) problems.
  • Block B (Resource Allocation): Once the "pathway" is set by Block A, Block B determines the optimal time for harvesting energy, the power for uploading/downloading data, and the local computation frequency for each device. This is solved using One-Dimensional Search Algorithms (ODSA) and the Bisection Algorithm (BA).

Crucially, they compare two STAR-RIS operating protocols: Energy Splitting (ES), where elements reflect and transmit simultaneously, and Time Switching (TS), where the surface alternates between reflecting and transmitting in different time slots.

Datasets

This study is based on numerical simulations rather than a static dataset. The authors modeled a network with:

  • A multi-antenna Access Point (AP).
  • A STAR-RIS with 60 elements.
  • Two groups of single-antenna IoT users (Group A in the transmission region, Group B in the reflection region).
  • Channel Models: They used Rician fading to account for both line-of-sight and non-line-of-sight components, with path loss exponents ranging from 3 to 3.5.

Key Insights and Ideas

  • TS is the Winner: The most significant "aha moment" is that Time Switching (TS) mode consistently outperforms Energy Splitting (ES) in terms of energy efficiency. This is because TS allows for more flexible phase-shift adaptation for each group of users and reduces the interference levels that occur when trying to serve everyone at the exact same time.
  • The "Full-Space" Advantage: Conventional RIS only covers a 180-degree "half-space." By using STAR-RIS, the system effectively doubles its reach, and the optimization ensures that users behind the surface aren't left in the dark.
  • Synergy of NOMA and FL: Using Non-Orthogonal Multiple Access (NOMA) allows multiple users to offload their local FL model updates simultaneously on the same frequency, which, when combined with the STAR-RIS's ability to "shape" the environment, drastically cuts down the time (and thus energy) spent on the uplink.

Key Results

  • Energy Efficiency: The TS-TS configuration (Time Switching in both uplink and downlink) achieved the lowest energy consumption across all tested scenarios.
  • Scaling: As the number of users ($K$) increases, the energy consumption of all schemes rises, but the TS-TS scheme maintains a significant lead over conventional RIS and ES-based schemes.
  • Hardware Impact: Increasing the number of STAR-RIS elements ($N$) or AP antennas ($M$) leads to a sharp decrease in total energy consumption, as it allows for much more precise energy and signal beamforming.
  • Convergence: The proposed BCD-based algorithm is computationally efficient, typically converging within fewer than 10 iterations.

Limitations and Future Work

While the results are promising, the authors acknowledge a few hurdles for real-world implementation:

  • Perfect CSI: The study assumes "Perfect Channel State Information" (CSI). In reality, knowing the exact state of every wireless link is difficult, especially with a smart surface in the mix.
  • Hardware Complexity: Implementing high-resolution phase shifters on a STAR-RIS is expensive and technically challenging.
  • Future Directions: The authors suggest exploring Rate-Splitting Multiple Access (RSMA) as an alternative to NOMA and investigating AI-based reinforcement learning to handle more dynamic, unpredictable environments where CSI is imperfect.

Summary updated

BibTeX
@article{alishahi2025efficient,
  author = {Mohammadhossein Alishahi and Ming Zeng and Paul Fortier and Omer Waqar and Muhammad Hanif and Dinh Thai Hoang and Diep N. Nguyen and Quoc-Viet Pham},
  title = {{Efficient STAR-RIS Mode for Energy Minimization in WPT-FL Networks With NOMA}},
  journal = {IEEE Trans. Communications},
  year = {2025},
  doi = {10.1109/tcomm.2025.3606641},
}
Figure