Rethinking WiFi-based Angle Estimation for Robust Passive Indoor Localization

Angle estimation is one of the commonly used techniques for WiFi-based indoor passive localization. However, existing works on WiFi-based passive angle estimation generally fail under two conditions. First, when the target exhibits minimal motion or moves in certain directions, existing solutions are unable to resolve the target reflection path from strong static paths and other dynamic paths. Second, when the target angle is large (i.e., exceeding ±45°), existing systems suffer from sharply increasing errors, which is traditionally attributed to the linear geometry of the antenna array. In this work, we provide the first in-depth analysis of the underlying causes behind these robustness issues and propose novel solutions to address them. Specifically, we propose using temporal signal differencing to eliminate strong static path components and introducing a dynamic differencing interval design to mitigate interference from other dynamic paths. To address the well-known issue of limited angle estimation range, we challenge the conventional wisdom and reveal that noise-induced phase wrapping, rather than array geometry, is the primary cause of large-angle errors. Correspondingly, we propose an iterative phase adjustment scheme that significantly expands the angle estimation range from ±45° to ±85°. We build a prototype system that achieves robust passive localization with an average angle error lower than 2°, outperforming the state-of-the-art with a mean error of 7°. Our angle estimation system works robustly even with only subtle respiratory motions. We believe this work represents a new milestone in WiFi-based angle estimation, moving one fundamental step toward high-precision and practical passive indoor localization.