Virtual Uplink Model in 6G & C-RAN

• Virtual Uplink Model is a mathematical abstraction that replaces complex wireless uplink processes with an equivalent channel for tractable analysis.
• It underpins system-level design in C-RANs, 6G SC³ loops, and hybrid networks by enabling techniques like noise matching and stochastic modeling.
• The model supports optimized resource allocation and performance prediction through convex optimization and standardized link-level simulations.

A virtual uplink model is a mathematical or algorithmic abstraction in which a wireless system’s uplink—whether a network of user devices, base stations, satellites, or sensing-control elements—is represented by an equivalent “virtual” channel. This abstraction enables analytic tractability, optimization, and system-level design by replacing the complexity of physical or architectural details with a stochastic or deterministic equivalence. Virtual uplink models are foundational across coordinated multipoint (CoMP), cloud radio access networks (C-RAN), 6G reflex-arc loops, and hybrid satellite-terrestrial networks.

1. Virtual Uplink Model in C-RANs and Cloud Processing

The canonical example of the virtual uplink model is the “virtual multiple-access channel” (VMAC) for uplink C-RAN. In this architecture, $K$ single-antenna users transmit to %%%%1%%%% single-antenna base stations (BSs), which are connected via finite-capacity, noiseless backhaul links to a central processor (CP) (Zhou et al., 2013). Each BS compresses its received signal using either distributed Wyner-Ziv coding (VMAC-WZ) or independent, single-user quantization (VMAC-SU), and forwards the quantized output to the CP.

After the CP receives all quantization codewords, the overall network is modeled as a VMAC: $$\hat{\mathbf{Y}} = \mathbf{H}\mathbf{X} + \mathbf{Z} + \mathbf{Q}$$ where $\mathbf{H}$ is the $L\times K$ matrix of channel gains, $\mathbf{Z}\sim\mathcal{CN}(0, \mathbf{N})$ accounts for receiver noise, and $\mathbf{Q}\sim\mathcal{CN}(0, \mathbf{Q})$ represents quantization noise.

The CP decodes users' messages treating the quantized received vectors as direct noisy observations, subject to the following information-theoretic constraints:

• Sum-rate achievable: For any subset $S\subset\{1,\ldots,K\}$,

$$\sum_{k\in S} R_k \leq I(\mathbf{X}_S ; \hat{\mathbf{Y}} | \mathbf{X}_{S^c}).$$

• Backhaul (rate-distortion) constraint:

$$I(\mathbf{Y}; \hat{\mathbf{Y}})\le \sum_{l=1}^L C_l.$$

Under these constraints, the system supports joint or successive decoding as if all remote user signals had been aggregated to the CP through a virtual MAC.

The optimal allocation of quantization noise ($\mathbf{Q}$) to maximize the weighted sum-rate under given backhaul limitations is nonconvex but can be addressed via alternating convex optimization with auxiliary variables. A significant result is that, in the high signal-to-quantization-noise-ratio (SQNR) regime, the quantization noises should be proportional to the background receiver noise ($q_\ell \propto \sigma_\ell^2$), and this “noise matching” is both near-optimal and guarantees a constant-gap approximation to the uplink capacity, even with simple per-node quantization (Zhou et al., 2013).

2. Virtual Uplink in Spatial and Stochastic Network Modeling

In hybrid satellite-terrestrial networks, the virtual uplink model is embedded within stochastic geometry frameworks for coverage probability analysis (Homssi et al., 2021). Terrestrial BSs, satellites, and users are modeled as spatial point processes—BSs as a planar PPP of density $\lambda_b$, satellites as points on a sphere approximated by an independent PPP with surface density $\lambda_s$, and user terminals as a planar PPP with density $\lambda_d$.

The core of the virtual uplink abstraction is that:

• Uplink delivery is considered successful if either the nearest BS or the nearest visible satellite can decode above a threshold SINR $\gamma_o$.
• Satellite and terrestrial link SNRs, interference statistics, and coverage probabilities are analytically tractable. Coverage is determined by $p_c = 1 - (1-p_s)(1-p_b)$, where $p_s$ and $p_b$ are the per-layer coverage probabilities, and the two layers are statistically decoupled except for their aggregate effect on the interference geometry.

Guided by this, tradeoff curves can be derived to meet a target uplink coverage reliability by balancing terrestrial densification (increasing $\lambda_b$) against satellite constellation scaling (increasing $N$) (Homssi et al., 2021).

3. Virtual User and Uplink-Downlink Coupling in 6G SC³ Loops

In the context of control-centric 6G networks supporting integrated sensing–communication–computing–control (SC³) loops, the virtual uplink model extends to treat physically and logically disjoint system elements as a single “virtual user” (Fang et al., 2024). Here, the sensor (plant), the edge/cloud processor, and the actuator are represented together:

• The uplink is redefined as the sensor-to-computing-center link;
• The downlink is the computing-center-to-actuator link;
• The entire chain (sensing, transmission, computation, actuation) over period $T$ is treated as a closed-loop virtual cycle subject to a linear quadratic regulator (LQR) objective.

The closed-loop entropy rate $D_{\mathrm{SC}^3} = \min\{\rho D_u, D_d\}$, where $\rho$ is a task-bit extraction ratio, and $D_u$, $D_d$ are bits transmittable in the uplink/downlink, emerges as the central system metric. Joint UL/DL/compute resource allocation is then posed as a convex optimization problem to maximize $D_{\mathrm{SC}^3}$ (subject to total time and bandwidth constraints), yielding closed-form bandwidth and time splits that enforce an information flow balance ($\rho D_u = D_d$) (Fang et al., 2024).

4. Implementation in Link-Level and Physical Layer Simulation

Virtual uplink models also underpin standardized link-level simulation, e.g., in the Vienna LTE-A Simulator (Zöchmann et al., 2015). There, the entire physical-layer transmission and reception chain for the LTE-A uplink is abstracted—the chain encompasses coding, mapping, DFT spreading (SC-FDMA), MIMO precoding, subcarrier mapping, channel modeling, CP addition/removal, and frequency- and time-domain processing.

The effective end-to-end input–output for the virtual uplink channel is: $$\hat{\mathbf{x}} = \mathbf{K}\mathbf{x} + \tilde{\mathbf{n}}$$ where $\mathbf{K}$ aggregates all deterministic and stochastic transformations, and $\tilde{\mathbf{n}}$ is aggregate noise post-processing. This abstraction enables the systematic evaluation of link adaptation, MIMO precoding, and equalization algorithms independent of implementation detail, and clarifies the distinction between user-layer and channel-layer model behaviors (Zöchmann et al., 2015).

5. Optimization Principles and Algorithmic Approaches

In both C-RAN and 6G SC³ virtual uplink scenarios, analytical convexity and structured optimization are central. For C-RAN:

• Alternating convex optimization coordinates quantization noise allocation with matrix inequality linearizations, yielding convergence to a stationary point (Zhou et al., 2013).
• A one-dimensional bisection over the “proportional noise” scalar ($q_\ell = \alpha \sigma_\ell^2$) offers near-identical performance to the full-dimensional method at drastically reduced complexity.

In 6G resource allocation:

• At optimality, time and bandwidth splits are balanced such that uplink and downlink links deliver exactly the needed task-relevant bits within the cycle time, maximizing the closed-loop entropy rate and thus minimizing LQR cost (Fang et al., 2024).

Key tradeoffs, such as between uplink and downlink bottlenecks or backhaul resource allocation versus quantization noise, are analytically explicit. System performance can be rapidly predicted or optimized via these canonical “virtual” model reductions.

6. Impact and System-Level Implications

Virtual uplink models offer the following system-level implications:

• They enable capacity-approaching cloud-processed networks via simple, analyzable settings, with robust guidelines such as “quantize proportional to receiver noise” for generalized C-RAN architectures (Zhou et al., 2013).
• Stochastic geometric virtual uplink models allow quantitative tradeoff analysis in heterogeneous, hybrid, or multi-layered networks, supporting joint optimization of terrestrial and satellite infrastructures under practical interference and resource constraints (Homssi et al., 2021).
• The virtual user and closed-loop entropy models unify cross-layer design for SC³ networks, making it possible to derive rigorous, closed-form optimality conditions for resource allocation that guarantee both communications and control-theoretic objectives (Fang et al., 2024).
• In link-level validation and standardization, the virtual uplink description decouples system blocks, enabling impartial inter-algorithm comparison and the quantitative prediction of performance as a function of channel, codebook, and MIMO strategies (Zöchmann et al., 2015).

7. Future Directions

The virtual uplink model continues to evolve:

• In C-RAN, extensions may encompass generalized quantization/compression techniques, non-Gaussian noise, and adaptive backhaul allocation under dynamic system loads.
• In hybrid and ultra-dense networks, virtual uplink abstractions are anticipated to support the design and analysis of networks beyond PPP approximations, modeling more realistic spatial distributions, mobility, and non-IID channel processes.
• For SC³ and 6G, further integration of communication–computation–control with stringent latency and reliability requirements is expected, leveraging the virtual uplink as a unifying interface between network-infrastructure optimization and real-time control performance.
• In simulation environments, richer virtual uplink models incorporating non-idealities, closed-loop adaptation policies, and high-fidelity hardware impairments are likely as system technologies advance.

These models thus provide the fundamental analytical backbone for next-generation wireless system design, analysis, optimization, and standardization.