This work-in-progress paper presents advancements in the acquisition function for processing GNSS signal snapshots, integrated within the System for Dynamic Repositioning (SyDR) framework. The primary objective of this research is to improve the accuracy and eficiency of GNSS signal processing, which is critical for applications such as navigation and positioning. The study addresses the challenges posed by complex wireless environments and the imperative need for energy-eficient solutions in embedded devices. Significant modifications were implemented in the existing positioning function, notably the removal of a Bayes classifier and the simplification of the acquisition function. The performance evaluation demonstrated substantial improvements in positioning accuracy, evidenced by a reduction in median errors and an increase of at least 5% in the percentage of errors below 200 meters for the test data. Despite these advancements, certain challenges persist, particularly concerning processing speed and performance issues encountered when integrating satellites from multiple GNSS systems. This paper contributes to the ongoing discourse on GNSS signal processing, ofering insights into the optimization of acquisition functions, and highlighting areas for future research to address the remaining challenges.
Over the past decade, Global Navigation Satellite Systems (GNSS) receivers have become a cornerstone
of modern industrial and person electronics, seamlessly integrated into a wide array of consumer and
industrial devices [
However, the evolving usage scenarios of GNSS introduce considerable challenges. Embedded
devices, particularly those operating in constrained environments, are often subject to adverse wireless
conditions, including signal obstruction, multipath propagation, and intentional or unintentional
interference. Moreover, unfavorable Dilution of Precision (DOP) metrics can significantly degrade
positioning accuracy. The design and optimization of GNSS receivers for such scenarios require
innovative strategies that ensure signal acquisition and tracking under limited resources. The
modernization of GNSS constellations, ofering enhanced signal structures, e.g., as additional frequencies,
improved coding schemes, and increased signal power, represents a step forward in mitigating
these limitations [
In response to these constraints, the research community has placed significant emphasis on developing novel algorithms to enhance signal processing eficiency. Numerous techniques have been proposed to incorporate the benefits of modernized signals into practical receiver designs. Yet, the trade-ofs between measurement accuracy, computational burden, and implementation feasibility remain non-trivial.
A critical challenge for embedded GNSS receivers is power consumption, particularly in
battery-operated or energy-constrained devices. According to the 2024 GNSS User Consultation
Platform, majority of respondents identified power consumption as a major design concern [
In this context, our ongoing research focuses on enhancing the acquisition function within the System for Dynamic Repositioning (SyDR) framework, first introduced in [ 13]. The objective is to integrate robust snapshot signal processing capabilities that are well-suited for constrained embedded platforms. Specifically, we target improvements in positioning accuracy, algorithmic eficiency, and support for multi-constellation scenarios.
The key contributions of this work include: (i) proposing the modification of the existing positioning function to support modern GNSS signals, (ii) initial simulation of representative test cases to validate performance under controlled conditions. Preliminary evaluations indicate measurable improvements in positioning accuracy. However, challenges remain, particularly in managing computational load and maintaining real-time responsiveness when handling signals from multiple GNSS systems, as well as factual integration in SyDR. Overall, this work contributes to the broader efort of developing energy-aware, high-performance GNSS solutions for next-generation IoT and cyber-physical systems.
The rest of the paper is organised as follows. First, we identify the main research objectives in Section 2. Selected numerical results are provided in Section 3. The last section concludes the paper.
A target platform is depicted in Figure 1. SyDR is an open-source SDR framework, implemented in Python and available via GitHub [14], specifically designed to facilitate benchmarking of GNSS algorithms. The design philosophy of SyDR is founded upon five fundamental principles, adapted from GNSS-SDR [15] and developed towards [13, 16], which define its intended functionality and usage for interoperability, usability reproducibility, openness, and eficiency.
Pedestrian
Vehicle
UAV Open-sky
Scenario Reference datasets
Urban canyoning
Forest
... ...
Software-Defined
Receiver Performance metrics
Algorithm ... Algorithm 1 n
New algorithm Processing chain
From its inception, SyDR was developed with algorithm benchmarking as a primary objective. It delivers a fully controllable and extensible framework that can be adapted to accommodate specific research requirements. In this respect, SyDR exhibits distinct diferences from previously introduced SDR platforms.
First and foremost, SyDR is designed exclusively for post-processing applications. It processes pre-recorded IQ samples while emulating the behavior of a real-time receiver. Post-processing enables the establishment of deterministic receiver behavior, which is essential for rigorous algorithm evaluation. Since GNSS data significantly afects algorithm performance, benchmarking must be conducted within a well-defined and repeatable environment to ensure that observed variations stem solely from the algorithm under test. The pseudo-real-time characteristics of the system allow for deeper insight into the virtual receiver’s behavior and the efect of algorithmic modifications across the signal processing pipeline.
Secondly, the framework emphasizes a high-level, modular development environment utilizing an open-source language. While many high-level SDR platforms have historically been implemented in proprietary environments such as Matlab, Python was chosen to maintain an open and community-driven development model while achieving satisfactory code eficiency. Modularity has often been a limiting factor in prior SDR platforms, which were typically tailored for evaluating a narrow set of algorithms rather than serving as comprehensive benchmarking tools.
Thirdly, the evaluation of energy consumption forms a key research direction for SyDR. Absolute power consumption figures are dificult to derive for specific hardware implementations when working with high-level software environments. In this context, algorithmic complexity serves as a practical proxy, ofering relative performance comparisons between algorithms. Although this metric does not directly reflect real energy usage, it provides meaningful insights into computational demands. Furthermore, by simulating the full receiver behavior, SyDR allows researchers to assess the impact of algorithmic changes on the overall system.
Ultimately, SyDR aims to serve as a reference platform for future research eforts in GNSS algorithm development, mitigating redundant implementation eforts. Although this work has primarily concentrated on the DSP aspects of the receiver, the framework’s modular design positions it as a promising foundation for further advancements in areas such as high-precision positioning and advanced signal processing techniques.
With an end-to-end virtual receiver like SyDR, it becomes possible to estimate the complexity of algorithms within the GNSS signal processing chain. While this estimation is inherently coarse, subject to variations stemming from code quality, operating system overhead, and other environmental factors, it still ofers a practical means of assessing relative algorithmic complexity through runtime measurements.
To enable this functionality, the time module from Python’s standard library has been employed, as recommended by the oficial documentation [ 17]. It is integrated into the codebase using a Python decorator , allowing benchmarking to be seamlessly added to any function. Owing to SyDR’s database architecture, every function wrapped with the benchmarking.time decorator logs its runtime as part of the intermediate results. This setup provides flexibility in measurement granularity, allowing users to identify the most time-consuming sections of processing with precision.
The primary development in this project involved improving the acquisition function to enhance snapshot positioning and integrating it into the broader SyDR framework. The numerical results obtained from this updated acquisition function demonstrated substantial improvements in positioning accuracy.
Only minimal changes were necessary to adapt the existing positioning function. A Bayes classifier was initially implemented to prioritize satellite selection based on their SNR values. However, it ofered no advantage over simpler SNR-based ranking and was thus removed. Additionally, support for processing multiple snapshots in a single call was removed to simplify functionality.
The original acquisition function in the library produced suboptimal results, even when supplied with accurate input data such as satellite visibility and expected Doppler values. Its reliance on complex high-dimensional vectorized operations made it dificult to interpret and debug. As a result, a complete reimplementation was undertaken. The new acquisition function combines vectorized 3D array operations with traditional for -loops, resulting in clearer and more maintainable code.
This revised implementation, using the same input parameters as the customized version, yielded markedly improved positioning performance. Specifically, the median position error decreased, and the percentage of errors under 200 meters increased by at least 5% on the test datasets. However, these gains came at the cost of reduced processing speed as the new function currently operates approximately ten times slower than the original.
Initial tests using TAU laboratory data yielded poor results due to an incorrect assumption about the data format. Upon discovering that the recordings were in IQ format (rather than real-valued), appropriate modifications were made. After these adjustments, the updated acquisition function performed well, albeit limited to one GNSS system at a time.
The simulation setup employed Spectracom (a.k.a., Orolia) and a USRP. Early issues involving the unavailability of trajectory and RINEX files [ 18] were resolved via firmware and software updates. Simulated datasets were created under the following configurations: • Static files : Contain signals from a single satellite. • Dynamic files : Include data from five satellites, with known PRN IDs. • Transmit Power and C/N0: Adjusted for each scenario. Note that power adjustments are possible only on the hardware simulator, not in StudioView.
• Recording Duration: One minute for static and five minutes for dynamic scenarios.
Sampling was conducted using 40 MHz and 16 MHz rates, with intermediate frequencies of 10 MHz and 4 MHz, respectively. Two processing phases were developed for pre-processing: • resampling : Implements zero-padding, upsampling, downsampling, resampling, and linear correlation via overlap-and-add.
• file_to_snapshots : Converts input recordings into snapshot data suitable for positioning.
These tools collectively enable the generation of consistent and controllable GNSS datasets, aiding in system testing and evaluation.
A total of 7 static and 4 dynamic simulated files (each with diferent 0 values) were recorded, dedicated to GNSS data. Visualization of these results is supported through the TAU data.ipynb notebook, which provides plots and statistics derived from results.npy , see, e.g., Table 1 and Figure 2. These visualizations include error metrics and per-mode performance summaries.
Scenario 5MHz_1bit_IQ_gain_25 5MHz_8bit_IQ_gain_25 10MHz_1bit_IQ_gain_25 10MHz_1bit_IQ_gain_25
Min East − 20.29 − 15.42 − 6.41 − 6.67
Max East 72324.01 19.70 7.97 6.21
Median East 0.72 − 1.68 − 0.80 − 1.03
RMS East 9047.28 6.93 3.88 3.28
SD East 8968.86 6.82 3.87 3.18
Error <50m, % 88 98 100 100
The results demonstrate substantial improvements in positioning accuracy, particularly in the later test cases. First, we observe a very high maximum error, suggesting significant outliers in the dataset, but the low median error indicates that most errors are negligible. Further cases show better performance, with maximum errors below 10 meters and median errors close to zero, indicating high accuracy. The high percentages of errors within 50 meters in these rows highlight the efectiveness of the updated acquisition function. However, the presence of extreme outliers underscores the need for further refinement and testing to ensure consistent performance across all scenarios.
The results align with expectations, i.e., positioning errors decrease with higher sampling rates and greater quantization depth. However, a significant issue arises when multiple GNSS systems are processed simultaneously. While the function performs well with a single GNSS system, accuracy degrades substantially when combining satellites from diferent systems. This issue also afects the ls-combo satellite selection mode, which evaluates all possible combinations of visible satellites and has been found to underperform relative to simpler selection schemes. This is a concerning outcome, as combining data from more visible satellites is theoretically expected to improve, not worsen, positioning accuracy. Overall, these findings highlight the potential of the proposed modifications to enhance positioning accuracy while also identifying areas for further optimization.
This research successfully enhanced the acquisition functionality for snapshot-based GNSS signal processing and integrated it into the SyDR framework. The improvements achieved, particularly in positioning accuracy, demonstrate the potential of the proposed modifications. However, challenges remain, especially when processing multiple GNSS constellations concurrently, which has led to degraded performance and requires further investigation.
The key findings from this work include improved positioning accuracy, with the updated acquisition function significantly reducing the median position error and increasing the percentage of errors under 200 meters. These improvements highlight the efectiveness of the new implementation in enhancing positioning accuracy. However, while the function performs well with a single GNSS system, accuracy degrades substantially when combining satellites from diferent systems. This issue also afects the ls-combo satellite selection mode, which underperforms relative to simpler selection schemes. This unexpected outcome indicates the need for further optimization and testing.
Future work will focus on addressing the identified limitations in multi-constellation processing and implementing SyDR on real hardware [16]. Specific areas for future research include ensuring the decoupling of acquisition pipelines for each GNSS system to improve performance when processing multiple constellations. Furthermore, resampling and signal preparation routines should be moved outside the acquisition block to reduce computational overhead and enhance overall system eficiency. Implementing SyDR on real hardware will validate the framework’s performance in practical scenarios and further refine the acquisition function.
In summary, this work has established a solid foundation for further research in GNSS signal processing using SyDR. The insights gained and the system improvements realized contribute meaningfully to the evolution of the SyDR platform and its applicability in real-world virtual receiver implementations.
The authors gratefully acknowledge funding from European Union’s Horizon 2020 Research and Innovation Programme under the Marie Skłodowska Curie grant agreement No. 956090 (APROPOS: Approximate Computing for Power and Energy Optimisation, http://www.apropos-itn.eu/)
During the preparation of this work, the authors used Microsoft Copilot in order to: styling, grammar and spelling checks. After using these tool, the authors reviewed and edited the content as needed and take full responsibility for the publication’s content. [12] H. M. Nguyen, Review of Of-loading Processing Strategies for GNSS Positioning: Theoretical Review and Visual Analysis Implementation, 2024. URL: https://trepo.tuni.fi/handle/10024/156652, examiners: Antoine Grenier, Simona Lohan. [13] A. Grenier, E. S. Lohan, A. Ometov, J. Nurmi, An Open-Source Software-Defined Receiver for GNSS Algorithms Benchmarking, in: Proc. of 14th International Congress on Ultra Modern Telecommunications and Control Systems and Workshops (ICUMT), IEEE, 2022, pp. 31–38. [14] A. Grenier, SyDR, [Online] https://github.com/aproposorg/sydr, 2023. [Accessed on June 10, 2025]. [15] C. Fernández–Prades, J. Arribas, P. Closas, C. Avilés, L. Esteve, GNSS-SDR: An Open Source Tool For Researchers and Developers, in: Proc. of the 24th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS 2011), Portland, OR, 2011, pp. 780–794. [16] A. Grenier, J. Lei, H. J. Damsgaard, E. S. Quintana-Ortí, A. Ometov, E. S. Lohan, J. Nurmi, Hard SyDR: A Benchmarking Environment for Global Navigation Satellite System Algorithms, Sensors 24 (2024) 409. [17] Python, Time Access and Conversions, https://docs.python.org/3/library/time.html, 2025. Last accessed: June 10, 2025. [18] International GNSS Service, RINEX Working Group, https://igs.org/wg/rinex/, 2025. Accessed: June 10, 2025.