Slow Movable Antenna System Design Based on Cell-Specific Long-Term Angular Power Spectrum

Technical Summary: Slow Movable Antenna System Design Based on Cell-Specific Long-Term Angular Power Spectrum

Problem Statement
Movable Antenna (MA) systems offer a promising paradigm for enhancing wireless performance by exploiting spatial degrees of freedom through flexible repositioning. However, existing designs predominantly rely on short-term, user-specific instantaneous or statistical Channel State Information (CSI). This reliance incurs substantial channel estimation overhead and excessive complexity due to the need for frequent antenna movements to track rapidly changing user locations or channel conditions. Furthermore, frequent mechanical repositioning increases energy consumption and wear, hindering practical deployment. While recent works have explored optimization based on statistical CSI, they often still depend on user-specific information (e.g., specific angles of departure or spatial covariance matrices for each user), which remains challenging to acquire in real-time and varies with user distribution.

Methodology
This paper proposes a "slow" antenna movement design framework that optimizes antenna positions over an extended timescale based solely on cell-level statistical channel information. The core methodology involves three main components:

1. Cell-Specific Statistical Channel Model: The authors develop a new channel model for MA-aided multiuser MISO systems. The cell is partitioned into subregions, where the downlink channel is modeled as a Gaussian vector under rich scattering assumptions. By weighting these subregion-specific distributions with the long-term user distribution, a cell-specific Gaussian mixture channel model is derived. From this, the Cell-Specific Angular Power Spectrum (APS) is defined, representing the averaged power responses of transmit paths across the entire cell, independent of specific user locations.

2. Covariance-Eigenvalues-Balancing Antenna Positions (CEBAP): To maximize the ergodic system utility (e.g., weighted sum rate or minimum SINR), the paper derives the CEBAP solution. Using Zero-Forcing (ZF) precoding and asymptotic analysis, the authors show that the optimal antenna positions asymptotically approximate the solution that equalizes the eigenvalues of the channel covariance matrix. This "eigenvalue balancing" effectively spreads the total spectrum power across the channel space, statistically reducing inter-user channel correlation. Crucially, the CEBAP solution depends only on the BS-side APS, not on instantaneous CSI or specific user locations.

3. Log-Barrier Penalized Optimization (LOBPO): To numerically solve the non-convex CEBAP optimization problem subject to physical constraints (moving region boundaries and minimum inter-antenna spacing), a low-complexity LOBPO method is proposed. This method incorporates non-convex constraints into log-barrier penalty functions and utilizes a gradient ascent algorithm with backtracking line search to iteratively converge to the optimal antenna positions.

Key Contributions

• Novel Channel Modeling: Development of a cell-specific Gaussian mixture channel model that captures both the scattering environment and long-term user distribution, leading to the definition of a cell-specific APS.
• Robust Optimization Framework: Derivation of the CEBAP solution, which maximizes ergodic utility by balancing channel covariance eigenvalues. This design is robust against varying system demands (different utility functions) and user distributions, requiring only the cell-specific APS.
• Efficient Algorithm: Proposal of the LOBPO method, a low-complexity numerical solver that handles the non-convex constraints of antenna positioning effectively.
• Performance Validation: Extensive simulations based on realistic urban layouts (Clementi, Singapore) and ray-tracing channel models demonstrate the efficacy of the proposed approach.

Results
Simulation results indicate that the proposed APS-based MA positioning consistently outperforms Fixed-Position Antenna (FPA) systems (both dense and sparse arrays) across various utility functions, including ergodic sum rate and minimum SINR.

• Performance Gains: The proposed design achieves significant performance improvements, particularly in moderately large antenna moving regions.
• Comparison with Benchmarks: The CEBAP approach closely approaches the performance upper bound achieved by antenna position optimization based on instantaneous CSI. It also outperforms optimization based on user-location-specific statistical CSI in several metrics (e.g., ergodic minimum SINR), likely due to the limitations of Monte Carlo sampling in the benchmark optimization.
• Robustness: The performance gains remain consistent across different transmit powers, Rician factors, and user distributions (uniform and Gaussian). The analysis shows that the benefits are most pronounced when the MA region is moderately large and the APS is moderately concentrated.

Significance and Claims
The paper claims that by shifting the optimization timescale from instantaneous to long-term and relying on cell-specific APS rather than user-specific CSI, the proposed design significantly alleviates the overhead associated with channel acquisition and antenna movement. This makes MA systems more practical for deployment. The authors emphasize that the CEBAP solution provides a utility-agnostic design that statistically reduces channel correlation, thereby improving system performance without the need for frequent reconfiguration or complex, real-time channel estimation. The work demonstrates that slow antenna movement based on long-term statistical knowledge can achieve performance levels comparable to those requiring instantaneous CSI, bridging the gap between theoretical gains and practical implementation constraints.