Ferkans — Interactive Telecom Tutor

From Statistical Models to Site-Specific Digital Twins

Wireless network design has traditionally relied on statistical channel models (Rayleigh, Rician, cluster-delay-line) that characterise the average behaviour of a propagation environment. These models are essential for standards and system-level simulation, but they cannot capture the site-specific details — the exact building geometry, material properties, foliage placement, and dynamic obstacles — that determine performance at a given location and time.

A wireless digital twin is a high-fidelity, continuously updated virtual replica of a physical wireless environment. It combines:

3D geometric model of the environment (from LiDAR, satellite imagery, or CAD building plans).
Ray-tracing propagation engine that computes site-specific path loss, multipath, and angular spectra.
Machine learning correction layer that calibrates ray-tracing predictions against real measurements, correcting for model inaccuracies (e.g., unknown material properties, diffuse scattering).
Real-time telemetry from the live network to keep the twin synchronised with the physical world.

The digital twin enables predictive network management: rather than reacting to measured KPIs, the network can simulate the effect of configuration changes (beam selection, power adjustment, handover thresholds) in the twin before applying them to the live network.

Definition:
ML-Calibrated Ray Tracing

Deterministic ray tracing propagates electromagnetic rays through a 3D environment model, accounting for reflection, refraction, diffraction, and scattering. For each Tx-Rx pair, the output is a set of multipath components (MPCs):

$h(\tau, \phi, \theta) = \sum_{\ell=1}^{L} \alpha_\ell \, \delta(\tau - \tau_\ell) \, \delta(\phi - \phi_\ell) \, \delta(\theta - \theta_\ell)$

where $(\alpha_\ell, \tau_\ell, \phi_\ell, \theta_\ell)$ are the complex gain, delay, azimuth AoA, and elevation AoA of path $\ell$ .

Limitations of pure ray tracing:

Requires accurate 3D geometry and material electromagnetic parameters (permittivity, conductivity), which are often unknown.
Computational cost scales with the number of interactions (reflections, diffractions) per ray — typically $\mathcal{O}(R \cdot I^3)$ for $R$ rays and $I$ maximum interactions.
Does not capture diffuse scattering and small-scale effects well.

ML-calibrated ray tracing addresses these limitations by training a neural network $f_\theta$ to predict the residual between ray-tracing predictions and field measurements:

$\hat{h}_{\text{corrected}} = h_{\text{RT}} + f_\theta(\mathbf{x}_{\text{env}}, \mathbf{x}_{\text{RT}})$

where $\mathbf{x}_{\text{env}}$ encodes environmental features (geometry, building type, foliage) and $\mathbf{x}_{\text{RT}}$ is the raw ray-tracing output. Training data comes from drive tests, MDT (minimisation of drive tests) reports, and crowd-sourced UE measurements.

State-of-the-art systems reduce the path-loss RMSE from 8 -- 12 dB (ray tracing alone) to 3 -- 5 dB (ML-calibrated), approaching the limit set by shadow fading variability.

Digital Twin Applications for 6G

A calibrated wireless digital twin unlocks several 6G capabilities:

Predictive beam management: Instead of sweeping beams reactively, the digital twin predicts the optimal beam for each UE based on its trajectory and the 3D environment. This can reduce beam-finding latency from 5 -- 20 ms (exhaustive sweep) to $< 1$ ms (direct prediction).
Proactive handover: Simulating the UE's path in the twin identifies the optimal handover point and target cell before the handover event, eliminating reactive measurements.
Network planning and optimisation: What-if analysis for new site placement, antenna tilt adjustment, or frequency refarming, all evaluated in the twin without affecting the live network.
Training data generation: The twin generates synthetic channel data for training AI/ML models (neural receivers, CSI compressors), augmenting scarce real-world measurements.
ISAC scene reconstruction: Sensing echoes from the communication signal are fused with the digital twin to detect and track objects (vehicles, pedestrians) in the environment — the intersection of digital twins and ISAC.

Definition:
Integrated Sensing and Communication (ISAC)

Integrated sensing and communication (ISAC) uses a single waveform and hardware platform to simultaneously perform wireless communication and radar-like sensing. The base station transmits an OFDM signal that:

Carries data to communication users (standard demodulation).
Illuminates the environment; echoes are received and processed to estimate target parameters (range, velocity, angle).

Sensing signal model: For a single OFDM symbol with $M$ subcarriers and subcarrier spacing $\Delta f$ , the echo from a point target at range $R$ and radial velocity $v$ is:

$Y[m, n] = \alpha \cdot X[m, n] \cdot e^{-j2\pi m \Delta f \cdot 2R/c} \cdot e^{j2\pi n T_{\text{sym}} \cdot 2v f_0/c} + N[m, n]$

where $m$ is the subcarrier index, $n$ is the OFDM symbol index, $X[m,n]$ is the known transmitted symbol, and $\alpha$ is the target reflectivity.

After dividing by $X[m,n]$ (known data), a 2D FFT across subcarriers (range) and symbols (Doppler) produces a range-Doppler map with resolutions:

$\Delta R = \frac{c}{2M\Delta f} = \frac{c}{2B}, \qquad \Delta v = \frac{c}{2 N_{\text{sym}} T_{\text{sym}} f_0}$

where $B = M\Delta f$ is the bandwidth and $N_{\text{sym}}$ is the number of symbols in the sensing frame.

The dual use of the OFDM waveform for sensing and communication creates a fundamental resource allocation trade-off: subcarriers/symbols allocated to sensing pilots improve range-Doppler resolution but reduce communication throughput. Joint waveform design that balances communication rate and sensing accuracy (the "rate-CRLB trade-off") is a central ISAC research problem.

ISAC and Full Duplex — The Monostatic Connection

Monostatic ISAC — where the sensing transmitter and receiver are co-located — requires the base station to receive echoes while transmitting. This is precisely the full-duplex problem of Section 33.4. The self-interference in this context is the direct path from Tx to Rx (through antenna coupling, reflections from nearby structures), while the desired signal is the weak target echo.

The SI cancellation chain (propagation + analog + digital) developed for full-duplex communication directly applies to monostatic ISAC. In fact, the requirements are even more stringent: radar targets at 100 m may produce echoes 140 -- 160 dB below the transmit power, requiring SI cancellation of 120+ dB.

Bistatic ISAC (separate Tx and Rx) avoids the SI problem but requires synchronisation between the transmitter and the sensing receiver, and provides different geometric coverage. Practical 6G ISAC will likely employ a mixture of monostatic and bistatic configurations.

ISAC Research Frontiers

ISAC is among the most actively studied 6G topics, with several research frontiers:

Joint beamforming design: How should the base station's spatial resources (beam directions, power allocation) be split between communication users and sensing targets? The Pareto frontier between communication sum-rate and sensing SINR is being characterised using multi-objective optimisation.
Waveform design: Beyond OFDM, dedicated ISAC waveforms (e.g., OFDM-chirp hybrids) can improve range-Doppler ambiguity properties while maintaining communication performance.
Sensing-aided communication: Information extracted from sensing (e.g., target position, velocity) can improve communication: predictive beam tracking, proactive blockage avoidance, and environment-aware scheduling.
Privacy and regulation: Sensing capabilities raise concerns about surveillance. Regulatory frameworks for cellular-based sensing are needed, along with technical privacy safeguards (e.g., sensing only aggregate statistics, not individual targets).
Standardisation: 3GPP has begun study items on ISAC for Release 19/20. The scope includes sensing use cases (indoor positioning, traffic monitoring, weather sensing), reference signal design, and performance metrics.

Example: ISAC Range and Velocity Resolution

A 6G base station operating at $f_0 = 28$ GHz uses an OFDM waveform with $B = 400$ MHz bandwidth and a sensing frame of $N_{\text{sym}} = 128$ OFDM symbols with $T_{\text{sym}} = 8.93\;\mu$ s (120 kHz SCS including CP).

(a) Compute the range resolution $\Delta R$ .

(b) Compute the velocity resolution $\Delta v$ .

(c) What is the maximum unambiguous range and velocity?

Solution

Range resolution

$\Delta R = \frac{c}{2B} = \frac{3 \times 10^8}{2 \times 400 \times 10^6} = 0.375\;\text{m}$ $

Sub-metre range resolution — comparable to automotive radar.

Velocity resolution

Total sensing frame duration: $T_f = N_{\text{sym}} \times T_{\text{sym}} = 128 \times 8.93 \times 10^{-6} = 1.143$ ms.

$\Delta v = \frac{c}{2 T_f f_0} = \frac{3 \times 10^8}{2 \times 1.143 \times 10^{-3} \times 28 \times 10^9} = 4.69\;\text{m/s} \approx 16.9\;\text{km/h}$

Maximum unambiguous range and velocity

$R_{\text{max}} = \frac{c}{2\Delta f} = \frac{3 \times 10^8}{2 \times 120 \times 10^3} = 1250\;\text{m}KATEXPLACEHOLDER0ENDv_{\text{max}} = \frac{c}{4 T_{\text{sym}} f_0} = \frac{3 \times 10^8}{4 \times 8.93 \times 10^{-6} \times 28 \times 10^9} = 300\;\text{m/s} = 1080\;\text{km/h}$ $The range and velocity resolutions are sufficient for urban sensing (vehicles, pedestrians), while the unambiguous limits cover practical scenarios.$ \blacksquare$

The Digital Twin as the Unifying 6G Framework

The digital twin concept ties together many of the technologies presented in this chapter:

Spectrum: Site-specific propagation prediction at FR3 and sub-THz frequencies, where statistical models are least accurate.
AI-native: The twin provides training data and a simulation environment for RL-based network management.
Near-field/XL-MIMO: Near-field channel models can be derived from ray tracing with spherical-wave propagation, avoiding the need for complex analytical models.
ISAC: Sensing echoes update the twin; the twin informs beam management and resource allocation.

In this view, the wireless digital twin is not merely a planning tool but a real-time operating system for the 6G network, continuously synchronised with the physical world and driving AI-based control decisions.

🎓CommIT Contribution(2025)

Deterministic-Random Tradeoff for ISAC

F. Liu, G. Caire — IEEE Transactions on Information Theory / IEEE Communications Society

Established the fundamental information-theoretic tradeoff between communication rate and sensing estimation accuracy (CRLB) in ISAC systems. The key insight is that communication requires random signalling (to carry information) while sensing benefits from deterministic waveforms (to minimise estimation variance). The optimal ISAC signal design lies on the Pareto frontier between these two objectives. This work provides the theoretical foundation for the joint waveform design problem discussed in this section and was awarded the 2025 IEEE ComSoc/IT Joint Paper Award.

isacinformation-theorywaveform-designparetoView Paper →

Integrated Sensing and Communication (ISAC)

A dual-functional system that uses a single waveform and hardware platform for both data communication and radar-like environment sensing. Range resolution $\Delta R = c/(2B)$ depends on bandwidth; velocity resolution depends on the sensing frame duration.

Related: Wireless Digital Twin

Wireless Digital Twin

A high-fidelity virtual replica of a physical wireless environment combining 3D geometry, ray-tracing propagation, ML calibration, and real-time network telemetry. Enables predictive network management, what-if analysis, and synthetic training data generation.

Quick Check

In an OFDM-based ISAC system, what determines the range resolution?

The carrier frequency $f_0$

The total signal bandwidth $B$ — range resolution is $\Delta R = c/(2B)$

The number of OFDM symbols in the sensing frame

The transmit power of the base station

Correction:

The total signal bandwidth

B

— range resolution is

\Delta R = c/(2B)

Correct. Range resolution depends only on the bandwidth: $\Delta R = c/(2B)$ . A wider bandwidth gives finer range resolution. This is the same fundamental relationship as in radar theory (the "range resolution equation"). The carrier frequency affects velocity resolution and maximum unambiguous velocity, but not range resolution.

Digital Twins, Ray Tracing, and ISAC