Interference and Multipath Resilient ToA Estimation

TL;DR

The paper addresses ToA estimation in interference-prone, multipath-rich indoor environments by introducing a GLRT-based, multi-antenna algorithm that jointly mitigates interference and resolves the first-arriving path with sub-$12$ ms latency using automatic differentiation (JAX). It abandons explicit model-order estimation and leverages a two-stage coarse/fine ToA strategy with a projected interference-subspace, validated on ray-traced indoor factory channels at 15 GHz. Results show significant improvements over correlation-based methods and JADE, especially under low SIR, while maintaining practical latency for synchronization and positioning tasks in dense deployments.

Abstract

We present a computationally-efficient algorithm for time-of-arrival (ToA) estimation that is robust under multipath propagation and strong interference. Our algorithm leverages multiple receive antennas to combine adaptive spatial filtering with autodifferentiation in order to super-resolve the tap of the first-arriving path at low computational complexity and without requiring model-order estimation. We use simulations with ray-traced indoor propagation channels to demonstrate significant performance improvements over conventional correlation-based ToA estimation methods and subspace techniques such as JADE.

Figures (4)

• Figure 1: Simulated indoor factory propagation scenario. (a) depicts the view from Wireless InSite and (b) the reconstructed view in MATLAB. (b) shows walls, floors, and simplified bounding boxes around the most prominent objects in the environment, as well as the transmitters in line-of-sight of one of the wall-mounted receivers, accounting for a limited $120^\circ$ FOV and occlusions.
• Figure 2: Detection performance of coarse ToA estimation stage measured by the area under the receiver operating characteristic (ROC) curve for different signal-to-interference ratios (SIRs). These results are for an SNR of $30\,\text{dB}$.
• Figure 3: Fine ToA estimation stage absolute timing error CDF for (a) a SIR of $-20\,\text{dB}$ and (b) a SIR of $\infty$. All results are shown for an SNR of $30\,\text{dB}$ and $I_\textnormal{F}=16$. The markers $\bullet$ denote the mean and $\blacksquare$ the median.
• Figure 4: Ablation studies for fine ToA estimation. (a) shows the impact of $I_\textnormal{F}$, fixing the SNR to $30\,\text{dB}$, and (b) shows the impact of the SNR, fixing $I_\text{F}=16$. The curves for FDNC and JADE are also shown for reference for an SNR of $30\,\text{dB}$.