Cell-Free MIMO Systems

• Cell-free MIMO systems are distributed multi-antenna wireless architectures that eliminate cell boundaries to provide uniform service and enhanced macro-diversity.
• Resource allocation in these systems leverages cooperative spatial processing and centralized power control to optimize both communication and surveillance scenarios.
• The architecture supports advanced applications such as broadband communications, URLLC, integrated sensing, and energy transfer while mitigating interference and shadow fading.

Cell-free MIMO systems are distributed multi-antenna wireless architectures in which a large number of geographically dispersed access points or monitoring nodes jointly serve or monitor all users or targets over the same time-frequency resource, without the notion of cell boundaries. In contrast to classical co-located or small-cell MIMO, cell-free deployments exploit spatial macrodiversity, uniform quality-of-service, and strong resilience against shadow fading, coverage holes, and interference. CF-MIMO has become a critical research paradigm in wireless networks, with applications ranging from broadband communications to URLLC, wireless surveillance, integrated sensing-and-communications, energy transfer, and unsourced random access.

1. Core Principles and Signal Model

A canonical cell-free MIMO (CF-MIMO) setup consists of $M$ distributed access points (APs) or monitoring nodes (MNs), each equipped with $N$ antennas, connected via fronthaul to a central processing unit (CPU). $K$ single-antenna user terminals (UEs), or in the surveillance context, untrusted transmitter–receiver pairs (UT–UR), are arbitrarily distributed in the coverage area. All APs/MNs jointly process uplink/downlink or surveillance signals using cooperative spatial processing.

Signal Model Overview:

• Uplink or Observation phase: Each UE/UT transmits to the network. Each AP/MN receives:

$$\mathbf{y}_m = \sum_{k=1}^{K} \mathbf{g}_{mk} s_k + \mathbf{w}_m$$

where $\mathbf{g}_{mk}\sim\mathcal{CN}(0, \beta_{mk} I)$ models the small–scale and large–scale fading.

• Downlink or Jamming/Transmission phase: Each AP transmits:

$\mathbf{x}_m = \sum_{k=1}^K \mathbf{w}_{mk} q_k$

Power or mode assignments may be determined centrally, with per-AP power constraints commonly enforced.

In surveillance settings, each MN can be in either “observing” or “jamming” mode, and the assignment is optimized for coverage or interception probability (Mobini et al., 2023). System operation often leverages MR, ZF, or MMSE beamforming/combining, with all APs (or a selected subset per user) jointly serving every terminal to maximize spatial diversity (Ngo et al., 2015, Zheng et al., 2023).

2. Resource Assignment, Optimization, and CSI Strategies

Mode Assignment and Power Control

Resource allocation in cell-free MIMO is inherently more complex due to the distributed hardware and cooperative nature of the network. For surveillance, observing-vs-jamming mode selection for MNs is performed using a greedy combinatorial algorithm that iteratively reassigns nodes to maximize the minimum monitoring success probability over all suspicious links (Mobini et al., 2023). For fixed mode assignment, long-term channel statistics suffice for per-MN power allocation via quasi-linear programming and linear program bisection, ensuring that power constraints and fairness metrics are tractable at scale.

For communication systems, max–min power control schemes maximize the minimum user rate through bisection-based feasibility checks on SINR constraints per AP (Ngo et al., 2015, Ngo et al., 2016). Rate-splitting, cluster-based user association, and user-centric clustering further extend scalability and robustness to CSI uncertainty (Flores et al., 2023, Zheng et al., 2023).

Channel State Information

CF-MIMO systems typically use TDD reciprocity and rely primarily on large-scale fading coefficients (long-term CSI) for network-wide decisions such as node assignment and beamforming codebook construction, minimizing small-scale CSI acquisition overhead. Centralized or distributed iterative methods (e.g., expectation propagation, ICD) further reduce pilot overhead and localize processing (He et al., 2021).

3. Performance Metrics and Analytical Results

Monitoring Success Probability

In cell-free surveillance architectures, “monitoring success probability” is defined as the probability that the SINR of the central monitor, derived from multi-MN observations and jamming assignments, exceeds the SINR at the target receiver:

$$P_{\mathrm{succ},k} = 1 - \exp\left(-\frac{\mathrm{SINR}_k^O \cdot \xi_k}{\beta_{kk}\rho_{UT}}\right)$$

where $\mathrm{SINR}_k^O$ is the monitor’s SINR, $\xi_k$ characterizes aggregate interference-plus-noise, and the formula is exact under Rayleigh fading and independence assumptions (Mobini et al., 2023).

Conventional Metrics

• Spectral efficiency ($\mathrm{SE}_k$): Achievable per-user (or per-link) rate derived via the “use-and-then-forget” bound, often based on closed-form SINRs that capture pathloss, shadowing, pilot contamination, and AP selection (Ngo et al., 2015, Zhang et al., 2017).
• Energy efficiency: Bit/Joule is modeled as system sum rate divided by aggregate AP, UE, and backhaul power consumption (Zhang et al., 2017, Zhang et al., 2020).
• Uniformity/fairness: Max–min power control leads to uniform QoS, with 95%-likely rates used to demonstrate “cell-edge” improvements (Ngo et al., 2016, Ngo et al., 2015).

Analytical Insights

• Distributing APs yields “macro-diversity” gains: SINR and cell-edge rates increase linearly with $N$0, surpassing small-cell architectures by up to 10–20× at the $N$1th percentile (Ngo et al., 2015, Ngo et al., 2016, Mobini et al., 2023).
• Closed-form hardware scaling laws: AP hardware quality may decrease as $N$2 increases (scaling with $N$3 for $N$4) without degrading per-user rates (Zhang et al., 2017).
• In cell-free surveillance, greedy mode assignment and long-term CSI-based power allocation can yield up to 11× gains in monitoring success probability over co-located architectures with the same aggregate hardware (Mobini et al., 2023).

4. Algorithmic Frameworks for Large-Scale and Low Complexity

Scalable Inference and Detection

Distributed expectation propagation combines local LMMSE updates at APs with centralized iterative MMSE, exchanging only extrinsic means and variances (“Gaussian messages”), achieving near-centralized performance without full exchange of small-scale CSI (He et al., 2021). For block-fading uplink, Neumann series approximations of Gram-matrix inverses enable $N$5 complexity for detection, close to optimal for practical truncation orders (Saray et al., 2020).

Scalable Random Access and User Scheduling

Cell-free architectures enable unsourced random access (URA) by partitioning detection workloads and fronthaul into per-AP subproblems; each AP performs OMP-based sparse recovery for signature detection and forwards a compressed statistic to the CPU, keeping per-AP complexity and fronthaul independent of the total number of active users. System capacity is increased by simply deploying more APs, embracing “scalable cell-free” as a design principle (Gkagkos et al., 2023).

Greedy and Bisection Algorithms

For surveillance and monitoring, greedy assignment of observing/jamming roles and bisection-based per-node power allocation lead to Pareto-optimal fairness without expensive combinatorial search (Mobini et al., 2023). This is tractable even with modest $N$6 due to reliance on long-term statistics and convex relaxation techniques.

5. Design Implications, Practical Considerations, and Comparative Analysis

Cell-Free vs. Co-Located and Small-Cell Baselines

• Cell-free architectures achieve orders-of-magnitude higher $N$7%-likely (cell-edge) throughput, monitoring success, and reliability compared to both co-located mMIMO and small-cell systems under identical AP and user densities (Ngo et al., 2015, Ngo et al., 2016, Mobini et al., 2023).
• Macro-diversity and the elimination of cell boundaries remove “cell-edge” bottlenecks and enable uniform service.
• Distributed deployments are more robust to correlated shadowing, hardware impairments, and adversarial attacks, as demonstrated in monitoring and SWIPT scenarios (Mobini et al., 2023, Alageli et al., 2019).

Hardware, CSI, and Resource Management

• Hardware Impairments: System-level SE is robust to low-cost, low-quality AP hardware as long as UE impairment is controlled and the number of APs is sufficiently large (Zhang et al., 2017).
• CSI Requirements: Knowledge of large-scale fading coefficients alone suffices for effective mode assignment and power control; full exchange of instantaneous small-scale CSI is not required, significantly reducing coordination and signaling overhead (Mobini et al., 2023, He et al., 2021).
• Full-duplex Emulation: Cooperative splitting of observing/jamming roles across MNs allows a half-duplex network to emulate full-duplex surveillance capabilities (Mobini et al., 2023).

Scalability, Fairness, and Deployment

• Greedy and distributed algorithms scale quadratically or linearly in $N$8 and $N$9 for typical settings, enabling real-world architectures with tens to hundreds of APs or monitoring nodes.
• Scalability is further enhanced by user-centric clustering and cluster-based rate-splitting to concentrate computation and fronthaul where most beneficial (Flores et al., 2023, Gkagkos et al., 2023).

6. Advanced Applications and Extensions

Surveillance and Jamming

CF-MIMO enables proactive wireless surveillance and jamming of untrusted links by splitting distributed MNs between observing and jamming modes, optimizing assignments and power via convex and greedy algorithms. The resulting system achieves high monitoring success probability, substantially outperforming co-located mMIMO in the same hardware regime, especially as the number of suspicious links grows (Mobini et al., 2023).

Integrated Sensing and Communication (ISAC)

Cell-free MIMO provides a natural platform for distributed ISAC, with architectures leveraging joint communication-and-sensing beamforming, multi-static distributed processing, and learning-based approaches (e.g., GNNs) for scalable joint optimization (Demirhan et al., 2024, Demirhan et al., 2023).

Energy Transfer and Secure Communication

The spatial macro-diversity and flexible per-AP power control in cell-free architectures yield superior energy efficiency and resilience in wireless power transfer (WPT), simultaneous wireless information and power transfer (SWIPT), and physical-layer security settings, particularly when robust to adversarial eavesdropping and active attacks (Alageli et al., 2019, Zhang et al., 2020).

---

References

• (Mobini et al., 2023) Cell-Free Massive MIMO Surveillance Systems
• (Ngo et al., 2015) Cell-Free Massive MIMO: Uniformly Great Service For Everyone
• (Ngo et al., 2016) Cell-Free Massive MIMO versus Small Cells
• (Zhang et al., 2017) Spectral and Energy Efficiency of Cell-Free Massive MIMO Systems with Hardware Impairments
• (Saray et al., 2020) A Low Complexity Space-Time Block Codes Detection for Cell-Free Massive MIMO Systems
• (He et al., 2021) Cell-Free Massive MIMO Detection: A Distributed Expectation Propagation Approach
• (Gkagkos et al., 2023) Scalable Cell-Free Massive MIMO Unsourced Random Access System
• (Flores et al., 2023) Clustered Cell-Free Multi-User MIMO Systems with Rate-Splitting