A Baseline Mobility-Aware IRS-Assisted Uplink Framework With Energy-Detection-Based Channel Allocation

1. Problem Statement

The paper addresses the complexity of managing uplink communications in low-power wireless networks (e.g., IoT) where users are mobile and channels are dynamic. Specifically, it tackles the challenge of integrating Intelligent Reflecting Surfaces (IRS) into a multi-user uplink system. Key difficulties include:

• Mobility: Tracking changing user geometries and time-varying fading.
• Resource Allocation: Efficiently assigning orthogonal channels to users to minimize interference while maximizing throughput.
• IRS Control: Determining which user to "focus" the IRS on and how to configure phase shifts with limited hardware precision (finite-bit quantization).
• Practicality: Existing models often rely on overly complex assumptions or lack a unified, self-contained framework that bridges theoretical physics-based modeling with implementable algorithms.

The goal is to establish a practical, consistent baseline framework that combines physics-based propagation modeling with adaptive scheduling and energy-detection-based channel allocation.

2. Methodology

The framework is built upon a simplified but explicit mathematical model and implemented via a MATLAB simulator.

A. System Model & Propagation

• Network Topology: $K$ single-antenna mobile nodes, a single-antenna Base Station (BS), and an IRS with $N$ elements.
• Channel Composition: The total channel $h_k(t)$ is the sum of a direct link and an IRS-assisted link.
• Path-Loss Modeling:
  • Uses a physics-based two-hop cascaded path-loss model.
  • The cascaded gain scales with $L_0^2$ (where $L_0 = (4\pi/\lambda)^2$), consistent with rigorous electromagnetic theory, rather than simplified empirical models.
  • Large-scale gains are distance-dependent; small-scale fading is modeled as independent complex Gaussian variables.
• Mobility & Fading:
  • Nodes move with constant velocity within a bounded region with reflective boundaries.
  • Fading Evolution: A hybrid approach is used. Small-scale coefficients are redrawn based on a coherence-time proxy derived from the maximum Doppler shift. Between redraws, phase evolution is approximated via a Doppler shift factor.

B. IRS Phase Control

• Focus Selection: The IRS focuses on a single "target" user per time slot.
• Phase Alignment: The simulator uses geometric phase alignment (compensating for path length) rather than perfect Channel State Information (CSI).
• Quantization: Phase shifts are quantized to $b$ bits (e.g., $b=3$, resulting in 8 levels).
• Power Scaling: The paper proves (Proposition 1) that under independence assumptions, the mean reflected power scales linearly with $N$ ($O(N)$), explicitly avoiding claims of deterministic $O(N^2)$ scaling which requires stronger coherence assumptions.

C. Adaptive Scheduling & Channel Allocation

• Focus User Selection:
  • Uses an inverse-rate priority weight: Users with lower recent average rates receive higher probabilities of being selected as the IRS focus target.
  • Policy: Round-robin during a warm-up period, followed by probabilistic selection based on a sliding window of past rates.
• Channel Allocation (Energy Detection):
  • Uses sequential channel allocation guided by energy detection.
  • Thresholding: An exact Chi-square ($\chi^2$) threshold is derived for the energy detector to maintain a target false-alarm probability ($P_{fa}$). For large sample sizes, a Gaussian approximation is also provided.
  • Algorithm: Users are assigned channels sequentially. A channel is selected if its estimated energy (noise + existing interference) is below the threshold; otherwise, the channel with the minimum energy is chosen.

3. Key Contributions

1. Unified Framework: A self-contained, implementation-aligned framework combining mobility, IRS control, and channel allocation.
2. Rigorous Path-Loss Formulation: Adoption of the correct $L_0^2$ cascaded scaling factor for two-hop IRS links, ensuring physical consistency.
3. Exact Detection Thresholds: Derivation and application of the exact Chi-square threshold for energy detection, moving beyond heuristic approximations.
4. Adaptive Policy Formalization: A formalized "inverse-rate" adaptive focusing policy that is fairness-aware by construction, though not claimed to be globally optimal.
5. Baseline Numerical Example: A seeded numerical simulation (MATLAB) demonstrating the framework's behavior, separating theoretical claims from specific numerical outputs.

4. Results (Numerical Example)

The paper presents a simulation with $K=10$ users, $N=64$ IRS elements, and $C=4$ channels over 200 time slots.

• Performance Heterogeneity: The system exhibits significant performance disparity.
  • Strong Users: Nodes with favorable geometry (e.g., Node 10) achieve high SINR (43 dB) and high rates (19.7 Mbps) but receive minimal IRS focus time (~1%).
  • Weak Users: Nodes with poor geometry (e.g., Node 4) often fall below the decode threshold (-10 dB) on average, despite receiving higher IRS focus time (12.5%).
• Fairness vs. Throughput:
  • The Jain's fairness index is low (0.366), indicating that while the scheduler attempts to prioritize weak users, the physical channel conditions dominate the outcome.
  • The adaptive rule successfully shifts focus time to lower-rate users, but this does not guarantee they can decode data if the channel is too poor.
• SINR Distribution: Average SINRs range from -10.79 dB to 43.43 dB, highlighting the extreme variability in mobility-aware IRS networks.

5. Significance and Future Directions

• Significance: This work provides a transparent, mathematically consistent baseline for future research. It avoids "black box" simulations by explicitly stating assumptions (e.g., independence of fading, geometric phase control) and separating theoretical derivations from simulation results. It serves as a reference for evaluating more complex mobility models or advanced scheduling policies.
• Limitations & Future Work:
  • The current mobility model is simplified (no full Jakes-type correlation).
  • The study relies on a single random seed; future work requires multi-seed averaging and confidence intervals.
  • Future extensions should explore learning-based scheduling, more realistic CSI acquisition, and control-delay effects in dynamically reconfigured IRSs.

In conclusion, the paper successfully demonstrates that while adaptive IRS focusing can prioritize weak users, physical channel limitations (path loss and fading) can still result in significant performance disparities, necessitating a robust baseline for further optimization.