Radio Network Digital Twin

The Radio Network Digital Twin (VectorizedChannelModel) simulates millimeter-wave (mmWave) 29 GHz propagation dynamics across 3D urban topographies.

Propagation Path Loss Model

Path loss \(\text{PL}(d)\) between transmitter \(i\) and receiver \(j\) at 3D distance \(d = \|\mathbf{p}_i - \mathbf{p}_j\|\) is modeled as:

\[\text{PL}(d) \; [\text{dB}] = \text{PL}_0 + 10 \cdot \eta \cdot \log_{10}\left(\frac{d}{d_0}\right) + X_\sigma\]

where:

\(\text{PL}_0\): Reference path loss at distance \(d_0 = 1\text{m}\) for carrier frequency \(f_c = 28\text{ GHz}\).
\(\eta\): Path loss exponent, taking distinct values based on geospatial LoS state:

\[\begin{split}\eta = \begin{cases} \eta_{\text{LoS}} \approx 2.0, & \text{if Line-of-Sight} \\ \eta_{\text{NLoS}} \approx 3.8, & \text{if Non-Line-of-Sight} \end{cases}\end{split}\]

\(X_\sigma \sim \mathcal{N}(0, \sigma^2)\): Log-normal shadow fading.

SINR & Shannon Data Rates

Signal-to-Interference-plus-Noise Ratio (SINR) for link \((i, j)\) with transmit power \(P_{\text{tx}}\):

\[\text{SINR}_{ij} = \frac{P_{\text{tx}} \cdot g_{ij} \cdot h_{ij}^{\text{PL}}}{\sigma_n^2 + \sum_{k \neq i} P_{\text{tx}} \cdot g_{kj} \cdot h_{kj}^{\text{PL}}}\]

Achievable data transmission rate \(R_{ij}\) follows Shannon channel capacity:

\[R_{ij} = B \cdot \log_2\left(1 + \text{SINR}_{ij}\right)\]

Vectorized Channel Computation

VectorizedChannelModel computes data rates for all links across thousands of environments simultaneously:

import torch
from urbanmarl.models.channel import VectorizedChannelModel

channel = VectorizedChannelModel(
    carrier_frequency=28e9,
    bandwidth=20e6,
    tx_power_dbm=30.0,
    noise_figure_db=9.0,
    device="cuda"
)

# Calculate data rates (B, N_tx, M_rx)
data_rates = channel.compute_data_rates(
    tx_pos=tx_positions,
    rx_pos=rx_positions,
    los_matrix=los_states,
)