AB-Sync: Attention-Based Slot-Level Clock Synchronization Method for UWB-TDOA Localization Networks

An attention-based approach to mitigate granularity mismatch in TDMA-based UWB-TDOA systems, improving localization accuracy without extra synchronization overhead.

Tianyi Lyu; Kefei Tian; Kangqiao Qin; Qingwen Liu; Mingqing Liu· 2026· DOI 10.48550/arXiv.2606.28087

The core problem

Ultra-wideband (UWB) time-difference-of-arrival (TDOA) localization networks are essential for high-update-rate indoor location services in IoT and cyber-physical systems. However, their accuracy critically depends on nanosecond-level clock synchronization among anchors. Existing wireless clock synchronization (WCS) methods typically estimate clock states at the synchronization-stage or interval level. In contrast, TDMA-based UWB-TDOA systems localize tags from blinks transmitted in discrete short slots within each synchronization stage. This granularity mismatch—between the synchronization-level clock estimates and the slot-level timestamps needed for tag localization—introduces residual TDOA errors. The paper identifies this mismatch as a key source of inaccuracy and proposes AB-Sync, an attention-based slot-level clock synchronization method that models the relationship between the slot-specific clock-speed ratio required by a target tag blink and neighboring clock-fluctuation observations. This enables tag-slot-level timestamp mapping without adding extra UWB synchronization messages.

The performance of AB-Sync was evaluated on a real UWB-TDOA testbed. Compared to Deferred+3S-KF, the leading low-overhead baseline, AB-Sync reduces the multi-anchor average TDOA ranging standard deviation (STD.V) by 9.4%. In terms of static localization accuracy, AB-Sync improves accuracy by 18.6% compared to the baseline. Furthermore, in a five-slot multi-tag experiment, AB-Sync consistently improves localization stability across all TDMA slots, reducing STD.V by 5.3% on average and up to 16.2% per slot. Importantly, these improvements are achieved with no extra UWB synchronization overhead. The results demonstrate that AB-Sync effectively mitigates the granularity mismatch and enhances localization performance in realistic scenarios.

Ultra-wideband (UWB) time-difference-of-arrival (TDOA) localization networks are essential for high-update-rate indoor location services in IoT and cyber-physical systems. However, their accuracy critically depends on nanosecond-level clock synchronization among anchors. Existing wireless clock synchronization (WCS) methods typically estimate clock states at the synchronization-stage or interval level. In contrast, TDMA-based UWB-TDOA systems localize tags from blinks transmitted in discrete short slots within each synchronization stage. This granularity mismatch—between the synchronization-level clock estimates and the slot-level timestamps needed for tag localization—introduces residual TDOA errors. The paper identifies this mismatch as a key source of inaccuracy and proposes AB-Sync, an attention-based slot-level clock synchronization method that models the relationship between the slot-specific clock-speed ratio required by a target tag blink and neighboring clock-fluctuation observations. This enables tag-slot-level timestamp mapping without adding extra UWB synchronization messages.

AB-Sync leverages an attention mechanism to predict the clock-speed ratio for each tag blink slot based on neighboring clock-fluctuation observations. The method operates within the existing TDMA framework, requiring no additional UWB synchronization messages. The core idea is to learn a mapping from observed clock fluctuations at nearby slots to the required clock-speed ratio at the target slot. This is formulated as an attention-based regression problem, where the attention weights dynamically emphasize relevant neighboring observations. The model is trained on data collected from a real UWB-TDOA testbed. The attention mechanism allows the model to focus on the most informative neighboring slots, effectively capturing the temporal correlation of clock fluctuations. The output is a slot-specific clock-speed ratio that is used to correct the timestamp of the tag blink, thereby reducing TDOA errors. The method is designed to be lightweight and compatible with existing TDMA schedules, ensuring no extra synchronization overhead.

The results indicate that the granularity mismatch between synchronization-level clock estimation and slot-level tag localization is a significant source of error in UWB-TDOA systems. By addressing this mismatch through attention-based slot-level clock synchronization, AB-Sync achieves substantial improvements in both ranging and localization accuracy. The method's ability to operate without additional synchronization messages makes it practical for deployment in existing TDMA-based UWB networks. The attention mechanism effectively captures the temporal dependencies of clock fluctuations, allowing accurate prediction of the required clock-speed ratio for each tag blink slot. The consistent improvements across multiple slots and tags suggest that AB-Sync is robust and generalizable. Future work could explore extending the approach to other wireless localization technologies and further optimizing the attention model for resource-constrained devices. Overall, AB-Sync presents a promising solution for enhancing the accuracy of UWB-TDOA localization networks in IoT and cyber-physical applications.

CS practitioners and researchers

Opening member content…