Abstract
The paper proposes a new method for the analysis and evaluation of the Quality of Service (QoS) in a mobile Internet access scenario. In particular, the paper proposes a throughput evaluation method based on the PathChirp algorithm. The end-to-end bandwidth was estimated by means of the Self Loading of Periodic Streams (SLoPS) technique. The obtained measurements were then analyzed by estimating the degree of correlation with other parameters that characterize the data transmission such as power, round trip time, etc. Finally, in order to have greater spatial resolution performance guaranteed by an Internet service provider, a 3D reconstruction method based on using drones is proposed and some preliminary results are discussed.
Introduction
Measuring mobile network performance has traditionally relied on "drive testing"—driving a vehicle equipped with sensors along specific routes. While effective for mapping coverage along roads, this method has two major flaws: it is restricted to ground level and cannot access areas unreachable by cars, such as steep terrain or large private complexes.
Furthermore, the standard way to measure throughput—downloading a file via FTP—presents a significant challenge. Small files provide quick but inaccurate results, while large files provide accuracy but take too long to download, especially in low-bandwidth areas. This measurement latency is a critical bottleneck for drones, which have limited battery life and must take multiple readings across a 3D coordinate grid.
Methodology
The researchers proposed a two-pronged solution: a high-mobility hardware prototype and a smarter bandwidth estimation algorithm.
Hardware Prototype
The team developed a custom sensor package using an ALIX2D2 system board, a 500 MHz CPU, and a USB internet key running on the ZeroShell OS. This package is lightweight enough to be carried by a civilian drone (under 10kg), enabling high-mobility 3D data collection.
PathChirp and SLoPS
Instead of traditional FTP, the system utilizes the Self Loading of Periodic Streams (SLoPS) technique, specifically implementing the PathChirp algorithm. Unlike standard methods that send equal-sized bursts of data, PathChirp sends "chirps"—trains of packets where the interval between packets gradually decreases. By measuring the One Way Delay (OWD) of these packets, the system can identify exactly when the transmission rate exceeds the available bandwidth, as the delay starts to increase due to network congestion. This allows for a rapid probe of network speed without transferring massive files.
Experimental Setup
Field tests were conducted in Catania, Italy, across 20 distinct test points in three scenarios:
- High-density urban area: High signal-to-noise ratio.
- Standard urban area: Variable signal strength.
- Suburban/Extra-urban area: Poor signal strength.
Results were compared against a "ground truth" established by a proprietary FTP-based measurement script.
Analysis and Results
The PathChirp algorithm significantly outperformed the alternative PathLoad algorithm in every test. For the 20 test points, PathChirp had an average percentage error of 21% compared to the FTP baseline, while PathLoad exhibited a much higher 56% error.
Key Insights
- Convergence Speed: PathChirp reached a stable estimate much faster than older algorithms. This efficiency allows a drone to capture a high-resolution 3D "heat map" of internet quality during a single flight.
- The RSSI Paradox: In crowded urban areas, high signal strength (RSSI) had almost zero correlation with internet speed, as network congestion was the primary bottleneck. In suburban areas, however, the correlation was nearly perfect.
- Spatial Resolution: Using drones allowed the identification of QoS changes with altitude—data invisible to traditional car-based testing but vital for future smart city infrastructure.
Conclusion
The PathChirp method enables accurate throughput estimation within a 25-second window, a necessity for drone-based sampling. While current battery technology limits flight times, future work will focus on optimizing flight paths to maximize spatial resolution. There is also a need to investigate other variables influencing QoS in high-density areas that current RSSI-focused models do not fully capture.
Summary updated