Radio Stripe Systems Overview

• Radio stripe systems are distributed antenna infrastructures that integrate active processing units along a shared cable, enabling scalable and cost-efficient cell-free massive MIMO and ISAC architectures.
• They employ both sequential and parallel signal processing methods to optimize latency and spectral efficiency, utilizing techniques such as MR, I-MR, and D-RZF.
• Practical designs address finite fronthaul capacity, AP selection, near-field RF wireless power transfer, and hardware-aware simulation to balance performance with deployment constraints.

Searching arXiv for the cited radio stripe papers and closely related work. Radio stripe systems are distributed antenna infrastructures in which antenna elements or active Antenna Processing Units (APUs) are integrated along a shared cable or serial fronthaul bus that provides power, synchronization, and data, enabling cable-like realizations of cell-free massive MIMO, indoor RF wireless power transfer (WPT), distributed integrated sensing and communications (ISAC), and related architectures (Ma et al., 2020, Azarbahram et al., 2023, Rosabal et al., 10 Apr 2026). Across this literature, radio stripes are treated as a cost-efficient and scalable way to deploy a distributed multi-antenna system, while preserving the possibility of local processing, short user-to-antenna distances, and distributed spatial diversity (Azarbahram et al., 2023).

1. Architectural definition and physical realizations

In cell-free massive MIMO, radio stripes are described as implementations in which $L$ access points (APs), each with $N$ antennas, are deployed along a stripe and share a single front-haul cable; the cable delivers power, synchronization, and data (Ma et al., 2020). In distributed ISAC, the physical realization is a series of distributed APUs embedded along building infrastructure such as walls or ceilings and interconnected via a serial fronthaul bus, with each APU equipped with $M$ antenna elements (Rosabal et al., 10 Apr 2026). In indoor RF-WPT, a radio stripe consists of a sequence of antenna elements spaced at fixed intervals $\kappa$ and typically mounted along the ceiling throughout the deployment area (Azarbahram et al., 2023). In the finite-capacity uplink architecture, each APU is connected to distributed, single-antenna APs integrated into the bus itself, exploiting macro-diversity more effectively (Chiotis et al., 2022). In the RS-GPA framework, radio stripes are used as active, individually controllable APU platforms along a shared cable, where selected APUs act as discrete and controllable radiation or reception points for location-flexible wireless access (Xu et al., 18 Jun 2026).

This suggests that “radio stripe system” denotes an architectural family rather than a single fixed embodiment. The common structural features are linear or quasi-linear physical integration, distributed transceiver functionality, and a serial interconnection fabric that couples radio processing to installation simplicity. The resulting design space spans centralized and distributed baseband strategies, full activation and sparse activation, and communication-only as well as communication-sensing or power-transfer operation.

2. Parallel and sequential signal processing in cell-free operation

A central theme in radio stripe research is the contrast between sequential processing and parallel processing. In the 2020 cell-free massive MIMO formulation, traditional processing such as the N-LMMSE scheme is sequential, so APs must wait for upstream neighbor processing before acting; latency therefore grows with the AP count $L$ (Ma et al., 2020). The proposed parallel schemes move the major computations to all APs simultaneously and serialize only simple addition across AP results as they propagate along the stripe, so total processing time equals that of one AP (Ma et al., 2020).

Within that framework, the basic uplink signal model uses the block-fading structure with $t_c$ symbols per coherence block, $t_p$ for pilots, and $t_c-t_p$ for data. The received data signal at AP $l$ is

$$\mathbf{y}_l = \sum_{k=1}^K \sqrt{p_k}\mathbf{h}_{k,l}s_k + \mathbf{n}_l,$$

and the per-user spectral efficiency is

$N$0

(Ma et al., 2020). Maximum ratio (MR) processing is fully parallelizable and has the lowest front-haul load, but its performance is poor under inter-user interference and near-far conditions (Ma et al., 2020). Interference-aware MR (I-MR) introduces weights based on large-scale interference statistics,

$N$1

and can be combined with a user-centric strategy that selects a subset of APs for each user (Ma et al., 2020). Distributed regularized zero-forcing (D-RZF), derived from LMMSE, uses parallel local projections and a central RZF step, and simulations show that it matches centralized LMMSE in both uncorrelated and correlated channels while keeping per-AP computational complexity close to MR (Ma et al., 2020).

The literature therefore frames radio stripes not only as a placement concept but also as a processing topology. The stripe enables local computation, serial aggregation, and reduced latency, but the specific performance-complexity trade-off depends strongly on whether the system adopts MR, interference-aware variants, distributed RZF, or more centralized inference.

3. Finite-capacity fronthaul, AP selection, and user-centric cooperation

When fronthaul or stripe capacity is finite, radio stripe design becomes a joint problem in quantization, cooperation clustering, and AP-UE association. The finite-capacity uplink analysis models quantized forwarding of pilots, data, soft estimates, and side information under a total capacity budget per coherence interval (Chiotis et al., 2022). The quantization error variance is

$N$2

where $N$3 is the number of quantization bits per symbol (Chiotis et al., 2022). The key algorithmic idea is Compare-and-Forward (CnF): at each APU, a normalized LMMSE soft estimate for user $N$4 is compared with the forwarded estimate from the previous APU, and the APU forwards whichever option yields better quantization-aware SINR. This induces a Dynamic Cooperation Cluster for each user and yields a user-centric radio stripe network approach (Chiotis et al., 2022). The paper states that, under finite capacity constraints, this solution can guarantee better uplink spectral efficiency than existing radio stripe architectures, especially when system size increases (Chiotis et al., 2022).

A complementary line of work studies AP-UE allocation as a discrete optimization problem. The AP-UE allocation matrix $N$5 determines which AP antennas serve which UEs, and the optimization objective is the sum uplink spectral efficiency at the last AP or CPU,

$N$6

with related sequential and parallel local objectives for distributed schemes (Conceição et al., 2024). The paper evaluates centralized maximum ratio combining (CMRC), centralized optimal sequence linear processing (COSLP), sequential MRC (SMRC), and parallel MRC (PMRC), and solves AP selection with a low-complexity and adaptive genetic algorithm (Conceição et al., 2024). COSLP exhibits the best spectral-efficiency performance at the expense of high computational complexity and fronthaul signalling, whereas SMRC and PMRC improve computational complexity and convergence speed relative to CMRC (Conceição et al., 2024). Reusing the AP-UE allocation matrix from a previous network state as an initial solution can significantly boost the optimization performance of the GA-based AP selection scheme in some user-addition and user-removal cases (Conceição et al., 2024).

A recurrent misconception is that radio stripes intrinsically require globally centralized processing. The cited results show that this is not the only viable regime: dynamic cooperation clustering, sequential local selection, and parallel local optimization all emerge as native stripe-compatible mechanisms, each with different fronthaul and convergence properties.

4. Near-field RF wireless power transfer and deployment optimization

In RF-WPT, radio stripes are used as distributed transmit apertures for indoor energy delivery to hotspots, defined as spatial zones with higher user density or consistent energy requirements (Azarbahram et al., 29 Aug 2025). The 2023 formulation considers $N$7 antenna elements with locations $N$8 and $N$9 energy hotspots with centers $M$0, under a near-field line-of-sight channel

$M$1

where $M$2 absorbs path loss and the antenna radiation pattern (Azarbahram et al., 2023). The received power at hotspot $M$3 is

$M$4

and the deployment objective is to maximize the minimum normalized received power across hotspots, subject to transmit-power, inter-element distance, and overall stripe-length constraints (Azarbahram et al., 2023). Maximum Ratio Transmission is used for practical tractability, leading to closed-form lower bounds and signomial/geometric-programming-based optimization (Azarbahram et al., 2023).

The 2025 extension generalizes this setup to multiple radio stripes and formulates a joint clustering and radio stripe deployment problem that aims to maximize the minimum received power across all hotspots (Azarbahram et al., 29 Aug 2025). The problem is decoupled into clustering and antenna element placement. The clustering stage uses the loss metric

$M$5

which reflects near-field path loss, with $M$6 the stripe center and $M$7 the hotspot (Azarbahram et al., 29 Aug 2025). The deployment stage then optimizes element locations $M$8, powers $M$9, and assignments $\kappa$0 in a fairness-oriented objective of the form

$\kappa$1

(Azarbahram et al., 29 Aug 2025).

Four deployment algorithms are reported: a fairness-aware clustering alternating-optimization method (FAC-AO), SGP-based deployment, SCA-based deployment, and a heuristic SGP mapping step that adjusts element positions to satisfy inter-element constraints (Azarbahram et al., 29 Aug 2025). The same work also studies two shape-constrained realizations. In regular polygon deployment, antenna elements are arranged as vertices of a regular polygon in the ceiling plane and optimization focuses mainly on the polygon center; in straight-line deployment, all elements are arranged in a straight line and optimization is over center position and rotation angle (Azarbahram et al., 29 Aug 2025). Numerical results show that Chebyshev initialization significantly outperforms random initialization in clustering, that optimized deployments consistently outperform baseline benchmarks across a wide range of frequencies and radio stripe lengths, and that polygon-shaped deployment achieves better performance than other approaches in most typical scenarios (Azarbahram et al., 29 Aug 2025). The line-shaped deployment demonstrates an advantage under high boresight gain settings, benefiting from increased spatial diversity and broader angular coverage (Azarbahram et al., 29 Aug 2025).

Two further findings are repeatedly emphasized in the WPT literature. First, increasing the stripe length can enhance performance (Azarbahram et al., 2023). Second, increasing the system frequency may degrade performance because increased channel loss can dominate the benefit of more elements within a fixed length (Azarbahram et al., 2023). These results caution against the simplistic view that larger frequency or larger aperture is automatically beneficial in near-field WPT.

5. Reconfigurable radio stripes for integrated sensing and communications

The ISAC formulation treats radio stripes as low-complexity distributed infrastructures in which homogeneous APUs can be dynamically configured either for communication or for sensing (Rosabal et al., 10 Apr 2026). At a given time, the APUs are partitioned into two disjoint sets: communication APUs $\kappa$2, which jointly serve $\kappa$3 single-antenna devices using OFDMA, and sensing APUs $\kappa$4, which acquire and process signals reflected from the environment due to communication transmissions (Rosabal et al., 10 Apr 2026). Because the APUs are homogeneous, each unit can be reassigned dynamically to optimize the sensing-communication trade-off (Rosabal et al., 10 Apr 2026).

For downlink ISAC, the $\kappa$5-th OFDMA subcarrier signal is

$\kappa$6

and the user receives $\kappa$7 (Rosabal et al., 10 Apr 2026). The sensing APUs observe multi-static echoes,

$\kappa$8

with $\kappa$9 formed by the superposition of reflections from multiple targets (Rosabal et al., 10 Apr 2026). Since target localization is non-convex in target positions, the service area is discretized into grid points, yielding a sparse linear model $L$0, followed by a consensus-ADMM formulation

$L$1

(Rosabal et al., 10 Apr 2026). Multiple sensing-communication assignments can then be fused through normalized global images weighted by normalized primal residuals (Rosabal et al., 10 Apr 2026).

The reported trade-offs are explicit. Increasing the number of devices and sensing APUs boosts sensing precision at the expense of degrading the sum rate, while the sum rate remains constant for a given number of communication APUs regardless of their positions (Rosabal et al., 10 Apr 2026). Changing the number of antennas has a non-monotonic impact on sensing performance because of the trade-off between array gain and illumination uniformity (Rosabal et al., 10 Apr 2026). This directly counters the common assumption that more antennas monotonically improve all system functions.

6. Radio stripes as generalized pinching-antenna realizations

The RS-GPA framework uses radio stripes as a practical realization of generalized pinching antennas, replacing passive dielectric-waveguide coupling with active APUs deployed along a shared cable for local transmission, reception, and signal processing (Xu et al., 18 Jun 2026). The channel between APU $L$2 and user $L$3 is modeled as a distance-dependent spherical wave,

$L$4

where $L$5 is the APU-user distance (Xu et al., 18 Jun 2026). This fine geometric dependence motivates sparse, location-aware activation.

For downlink transmission, the joint APU activation and beamforming problem minimizes total network power,

$L$6

subject to user SINR constraints, per-APU power constraints, and the requirement that at least $L$7 APUs be active (Xu et al., 18 Jun 2026). The paper develops a reweighted group-sparse beamforming algorithm by replacing the $L$8-type activation indicator with a reweighted $L$9-type penalty on row powers, solved as a reweighted SOCP (Xu et al., 18 Jun 2026). In the single-user case, the total power with the $t_c$0 closest APUs is

$t_c$1

and an additional APU should be activated only if the transmit-power saving exceeds the circuit-power cost (Xu et al., 18 Jun 2026). A geometry-guided low-complexity multiuser algorithm is then built around the score

$t_c$2

(Xu et al., 18 Jun 2026).

For uplink transmission, the framework formulates joint APU activation and user power control, with MMSE receive combiners for a given active set and a geometry-guided sparse activation design for the combinatorial part (Xu et al., 18 Jun 2026). Numerical results show that the RS-GPA framework substantially reduces total consumed power relative to benchmark schemes, while the geometry-guided algorithm achieves near-identical consumed-power performance to the group-sparse design with significantly lower runtime (Xu et al., 18 Jun 2026). A plausible implication is that radio stripes naturally support energy-aware sparse activation because their spatially distributed APUs can be selected according to user geometry rather than kept permanently active.

7. Hardware-aware simulation, evaluation methodology, and recurring design trade-offs

The open-source sub-THz radio-stripe simulator extends the literature from analytical models to waveform-level, hardware-aware emulation (Schepens et al., 16 Apr 2026). It models the full signal chain from CP-OFDM baseband generation in the central unit, through measurement-parameterized polymer microwave fiber and coupler links, to booster or active Radio Units with configurable nonlinearity, noise, in-phase and quadrature imbalance, oscillator phase noise, and carrier frequency offset (Schepens et al., 16 Apr 2026). Wireless propagation is available through deterministic, stochastic per-subcarrier, and site-specific ray-tracing channels generated with a companion Sionna ray-tracer module, and the simulator exports intermediate waveforms and metrics including normalized mean square error, signal-to-noise-and-distortion ratio, and bit error rate (Schepens et al., 16 Apr 2026).

The measurement and simulation toolchain reinforces several cross-cutting results already visible in the algorithmic literature. First, front-haul and hardware impairments can strongly shape end-to-end performance, so stripe evaluation cannot be reduced to idealized channel matrices alone (Schepens et al., 16 Apr 2026). Second, radio stripe performance is highly configuration dependent: polygon-shaped WPT layouts often outperform unconstrained or line-shaped alternatives, but line deployments can dominate under high boresight gain settings (Azarbahram et al., 29 Aug 2025); more sensing APUs improve localization but reduce rate (Rosabal et al., 10 Apr 2026); higher frequency can degrade WPT despite denser element packing (Azarbahram et al., 2023); and centralized algorithms can maximize spectral efficiency while remaining unattractive in runtime or fronthaul requirements (Conceição et al., 2024). Third, user-centric or geometry-guided control recurs across domains—dynamic cooperation clusters in finite-capacity uplink (Chiotis et al., 2022), AP selection by genetic optimization (Conceição et al., 2024), and sparse APU activation in RS-GPA (Xu et al., 18 Jun 2026).

Taken together, the literature presents radio stripe systems as a unifying distributed-wireless substrate rather than a single-purpose technology. Their distinguishing attribute is the tight coupling of cable-based physical integration with distributed spatial processing. The main research questions are therefore no longer limited to coverage and spectral efficiency; they also include deployment geometry, fronthaul quantization, sparse activation, sensing-communication role assignment, near-field focusing, and hardware-aware reproducibility.