Previous Abstract Return to Session B6 Next Abstract

Session B6: Emerging PNT Consumer Applications

Low-Power Positioning on Wearable: Integration of Snapshot GNSS and Neural PDR
Chin Lok Tsang, Shiyu Bai, Di Hai, Hoi-Wah Ng, Li-Ta Hsu, Department of Aeronautical and Aviation Engineering, The Hong Kong Polytechnic University
Location: Prince David Room
Date/Time: Thursday, Apr. 16, 1:49 p.m.

Outdoor positioning has become a critical function on wearable devices, allowing users to track their runs, hikes, and other outdoor activities [1-2]. Global Navigation Satellite Systems (GNSS) are typically used for outdoor positioning because they provide accurate position, velocity, and time (PVT) solutions anywhere in the world through absolute positioning. However, GNSS service consumes substantial power [3], posing significant challenges for power-constrained wearable devices that rely on small
This high power consumption stems from how conventional GNSS receivers operate. GNSS receivers generally acquire and track satellite signals channel by channel, with each tracking channel dedicated to processing signal from a single satellite. Each channel performs complex early-prompt-late (E-P-L) correlation operations continuously to maintain lock on its assigned satellite signal [4]. This continuous multi-channel operation makes GNSS receivers one of the most power-hungry components in wearable devices [5], often limiting how long positioning features can remain active before draining the battery.
In recent years, GNSS snapshot positioning (S-GNSS) has become popular again as a way to reduce power consumption in IoT applications such as wildlife tracking, asset tracking, and healthcare monitoring [6-7]. Unlike conventional GNSS receivers, S-GNSS receivers do not employ a tracking engine for continuous signal tracking. Instead, the receiver turns on periodically to capture a short batch of signal data, sometimes as brief as a few milliseconds [8]. Using this captured signal, the receiver then employs an open-loop batch processing architecture to estimate the signal code phases, which are used to initialize the snapshot navigation engine and compute the position fix. This method significantly reduces the complexity of GNSS baseband processing and lowers power consumption. However, rather than continuous tracking, the position fixs from S-GNSS are in discrete form. This fails to capture the actual path taken, especially around curves or turns, making the activity data less comprehensive. Speed calculations also become unreliable when based on sparse data points, making it harder to provide accurate real-time metrics like current pace or estimated arrival time.
Beyond S-GNSS, extensive research has recently focused on pedestrian dead reckoning (PDR) as a self-contained and low-power localization alternative. PDR estimates position by accumulating displacements using built-in IMU sensors on smartphones or wearable devices [9]. However, IMU-based systems provide only relative positioning, which cannot replace the absolute positioning capability of GNSS. Furthermore, PDR inherently suffers from drift because estimation errors accumulate over time, leading to substantial position deviations during long-term operation [10]. To address these limitations, GNSS-PDR fusion solutions using Extended Kalman Filters (EKF) have been proposed. These systems correct PDR drift by incorporating absolute positioning measurements from GNSS. However, GNSS-PDR fusion still requires continuous GNSS signal tracking, where complex E-P-L correlation operations are needed to generate GNSS measurements. Consequently, power consumption remains significant, which contradicts the original low-power positioning intention of PDR.
Recognizing this research gap in low-power positioning for wearables, this work proposes a snapshot GNSS-PDR (S-GNSS-PDR) fusion solution to address these challenges. To the authors' knowledge, no previous work has explored S-GNSS-PDR fusion. This novel approach enables the two systems to complement each other's shortcomings while maintaining low power consumption. PDR provides continuous position tracking, while the S-GNSS receiver periodically delivers discrete absolute position fixes that correct for PDR drift. Although the positioning accuracy may not be as accurate as conventional GNSS-based solutions, S-GNSS-PDR achieves a favorable balance between accuracy and power consumption. This is accomplished by minimizing power-hungry GNSS baseband processing and eliminating the need for prolonged signal tracking to maintain satellite lock.
In this work, L5-band-based S-GNSS is employed to address the inherent measurement challenges of snapshot receivers [11]. Since S-GNSS receivers operate without loop filters, their code phase measurements are relatively noisy. Conventional open-loop receivers typically use long integration times for signal noise suppression [12]. However, S-GNSS receivers capture only brief signal snapshots, making long coherent integration times unavailable for noise suppression. The L5 band offers two key advantages that compensate for this limitation. First, it provides higher received signal power. More significantly, its increased bandwidth substantially improves noise suppression and multipath mitigation capabilities, thereby enhancing the accuracy of code phase measurements. The estimated L5 signal code phase measurements from different satellites are then used to perform snapshot positioning. Since the precise signal transmission time is unknown, an additional coarse time state is added to the least squares estimation. This coarse time parameter adjusts the satellite positions iteratively until the least squares estimation converges. Once convergence is achieved, the position fix is used to correct or initialize the PDR estimate. In this work, the snapshot receiver activates periodically every 30 seconds, providing position fixes at 0.033 Hz. With the position being initialized by S-GNSS, the PDR would continue for the propagation to provide a position fix in a continuous form. For PDR implementation, this work adopts a neural inertial odometry approach [13,14]. Traditional PDR methods rely on step detection, step length estimation, and heading estimation to infer displacement at each step. These human-motion-model-based methods, however, degrade under diverse motion patterns and irregular arm movements in wearable scenarios. Instead of estimating step parameters, our method directly learns the relationship between IMU measurements and velocity. A motion-aware attitude estimation module and a lightweight neural network are integrated to accurately regress velocity while maintaining low computational burden for wearable [13], allowing S-GNSS updates to occur at lower rates to minimize power consumption while still providing effective drift correction. Regarding power consumption, S-GNSS-PDR's algorithm complexity and runtime required for processing will be compared with a GNSS-only solution to demonstrate its low-power property.
Significance of the work:
1. A novel S-GNSS-PDR fusion approach is proposed to enable low-power, continuous positioning with minimal drift over extended durations.
2. A state-of-the-art neural inertial odometry (NIO) framework is introduced to minimize the drift rate of PDR, which allows the S-GNSS update rate to be reduced to as low as 0.033 Hz.
3. The positioning performance of S-GNSS-PDR is evaluated through real-world experiments, with comparisons made against GNSS-only in SPP (GNSS-SPP) and PDR-only positioning methods.
4. Algorithm complexity and runtime comparison between S-GNSS-PDR and a GNSS-only solution to demonstrate S-GNSS-PDR's low-power property
Reference:
[1] Vyas, A., & Pal, S. (2023). Power saving approach of a smart watch for monitoring the heart rate of a runner. IEEE Transactions on Consumer Electronics, 69(3), 490-498.
[2] Birenboim, A., Dijst, M., Scheepers, F. E., Poelman, M. P., & Helbich, M. (2019). Wearables and location tracking technologies for mental-state sensing in outdoor environments. The Professional Geographer, 71(3), 449-461.
[3] Grenier, A., Lohan, E. S., Ometov, A., & Nurmi, J. (2023). A survey on low-power GNSS. IEEE Communications Surveys & Tutorials, 25(3), 1482-1509.
[4] Morton, Y. J., van Diggelen, F., Spilker Jr, J. J., Parkinson, B. W., Lo, S., & Gao, G. (Eds.). (2021). Position, navigation, and timing technologies in the 21st century: Integrated satellite navigation, sensor systems, and civil applications, volume 1. John Wiley & Sons.
[5]Sabola, V. L., Granados, G. S., Salcedo, J. A. L., & García-Molina, J. A. (2018). GNSS IoT positioning: From conventional sensors to a cloud-based solution. Inside GNSS.
[6] Beuchert, J., & Rogers, A. (2021, November). SnapperGpS: Algorithms for energy-efficient low-cost location estimation using GNSS signal snapshots. In Proceedings of the 19th ACM conference on embedded networked sensor systems (pp. 165-177).
[7] Lucas-Sabola, V., Seco-Granados, G., López-Salcedo, J. A., & García-Molina, J. A. (2018). Performance analysis of low-power GNSS positioning in IoT. Proc. NAVITEC, 1-4.
[8] Van Dierendonck, K., Al-Fanek, O., & Petovello, M. (2018). What is snapshot positioning and what advantages does it offer. Inside GNSS, 6, 28-33.
[9] Yan, D., Shi, C., & Li, T. (2022). An improved PDR system with accurate heading and step length estimation using handheld smartphone. The Journal of Navigation, 75(1), 141-159.
[10] Bai, S., Wen, W., Li, Y., Shi, C., & Hsu, L. T. (2024). Towards Persistent Spatial Awareness: A Review of Pedestrian Dead Reckoning-Centric Indoor Positioning with Smartphones. IEEE Transactions on Instrumentation and Measurement.
[11] Tsang, C. L., Ng, H.W., Luo, Y., Hai, D.,& Hsu, L. T. Ultra-low power GNSS L5 band snapshot positioning: A two step approach
[12] Tsang, C. L., Luo, Y., & Hsu, L. T. (2022). Long Coherent Open-Loop GPS L5Q Signal Positioning: A Case Study for an Urban Area in Hong Kong. In Proceedings of the 35th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2022) (pp. 2025-2041).
[13] Bai, S., Lyu, Y., Lyu, Z., Xu, R., Wang, X., & Wen, W. Watch Your Position: Neural Inertial Localization with a Single Wrist-worn Device.
[14] Bai, S., Wen, W., Su, D., & Hsu, L. T. (2025). Graph-based indoor 3d pedestrian location tracking with inertial-only perception. IEEE Transactions on Mobile Computing.

Previous Abstract Return to Session B6 Next Abstract