Calibrating Wireless Ray Tracing for Digital Twinning using Local Phase Error Estimates

TL;DR

This work tackles RT calibration for wireless DTs by addressing phase errors caused by small geometric mismatches. It introduces a phase-error-aware calibration scheme built on variational expectation maximization (VEM) with a von Mises prior on per-path phase errors, enabling a closed-form E-step and gradient-based M-step to jointly refine material parameters and per-path phase estimates. By leveraging differentiable RT (Sionna) and data-driven CFR observations, the method outperforms phase-oblivious and uniform phase-error baselines in both parametric accuracy and power prediction, particularly at lower bandwidths where phase information is crucial. The approach enhances DT fidelity and the reliability of RT-based data augmentation and prediction for wireless network design and operation.

Abstract

Embodying the principle of simulation intelligence, digital twin (DT) systems construct and maintain a high-fidelity virtual model of a physical system. This paper focuses on ray tracing (RT), which is widely seen as an enabling technology for DTs of the radio access network (RAN) segment of next-generation disaggregated wireless systems. RT makes it possible to simulate channel conditions, enabling data augmentation and prediction-based transmission. However, the effectiveness of RT hinges on the adaptation of the electromagnetic properties assumed by the RT to actual channel conditions, a process known as calibration. The main challenge of RT calibration is the fact that small discrepancies in the geometric model fed to the RT software hinder the accuracy of the predicted phases of the simulated propagation paths. Existing solutions to this problem either rely on the channel power profile, hence disregarding phase information, or they operate on the channel responses by assuming the simulated phases to be sufficiently accurate for calibration. This paper proposes a novel channel response-based scheme that, unlike the state of the art, estimates and compensates for the phase errors in the RT-generated channel responses. The proposed approach builds on the variational expectation maximization algorithm with a flexible choice of the prior phase-error distribution that bridges between a deterministic model with no phase errors and a stochastic model with uniform phase errors. The algorithm is computationally efficient, and is demonstrated, by leveraging the open-source differentiable RT software available within the Sionna library, to outperform existing methods in terms of the accuracy of RT predictions.

Figures (13)

• Figure 1: Taking as input the geometric properties of the scene, the electromagnetic material parameters $\theta$, and the coordinates $c$ of transmitter (Tx) and receiver (Rx), the ray tracer (RT) produces the features $R(c|\theta)$ of a number $P$ of propagation paths. Based on this information, the DT can obtain a model $H(c | \theta)$ of the channel conditions between transmitter and receiver. To keep a faithful representation of its physical twin (PT), during a calibration phase, the DT compares its model predictions $H(c | \theta)$ to measured channel realizations $\mathcal{D} = \{H_n, c_n\}_{n=1}^{N}$ in order to optimize the material parameters $\theta$. An accurate estimate of the ground-truth channel conditions can then be used to monitor the PT, simulate alternate scenarios (counterfactual analysis), or to optimize PT operations such as beamforming, with reduced requirements on pilot transmissions and channel measurements jiang2023digital. (The map and the 3D model in the the figure are built using geospatial data from the OpenStreetMap database OpenStreetMap.)
• Figure 2: Toy example that illustrates the limitations of existing power profile-based RT calibration methods, as well as of channel response-based schemes that disregard path phase errors. In this example, the two propagation paths between the transmitter (Tx) and the receiver (Rx) reflect on surfaces with ground-truth reflectance $r^{*}$ in (a), and calibrated reflectance $\hat{r}$ in the simulated scenario (b). Though the paths interfere constructively at the receiver under ground-truth conditions, they are predicted to interfere destructively in the simulated scenario due to an inaccuracy $\Delta d$ of the order of the carrier's wavelength $\lambda$ in the geometric model. This difference is illustrated as the blue dashed lines in (c) and (d), which represent the signed amplitudes of each path. Due to this model error, phase error-oblivious schemes provide inaccurate calibration results in both high and low-bandwidth regimes (e)-(f). Furthermore, the model error causes the RT-generated power-delay profiles in the low-bandwidth regime (d) to diverge from the ground-truth power-delay profile (c), causing inaccurate calibration also for power profile-based schemes.
• Figure 3: Illustration of the definitions of angle path parameters for planar antenna arrays. The red crosses represent the elements of each array. The angles and directions of the departing and arriving signals, as well as the positions of the array elements, are represented in the respective local coordinates.
• Figure 4: Von Mises probability density functions $\mathcal{VM}(0, \kappa)$ for different values of the concentration parameters $\kappa$.
• Figure 5: Relative power estimation errors at the position $c$ used to collect data for phase error-oblivious calibration (Sec. \ref{['subsec:peoc']}), uniform phase error calibration (Sec. \ref{['subsec:upec']}), and for the proposed phase error-aware calibration (Sec. \ref{['sec:peac']}) as a function of the available bandwidth $B$, with a signal-to-noise (SNR) ratio of $20$ dB. Lines represent the median error across ten independent channel observation and calibration runs. Shaded areas represent the first and third quartiles.