⚡ electrical engineering

Autonomous Sensing UAV for Accurate Multi-User Identification and Localization in LAWN

This paper presents and validates an autonomous, infrastructure-independent UAV framework that passively senses 5G uplink signals to simultaneously identify and localize multiple users with high accuracy, demonstrating its feasibility for enhancing situational awareness in the emerging Low Altitude Economy.

Original authors: Niccolò Paglierani, Francesco Linsalata, Vineeth Teeda, Davide Scazzoli, Maurizio Magarini

Published 2026-03-03

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Niccolò Paglierani, Francesco Linsalata, Vineeth Teeda, Davide Scazzoli, Maurizio Magarini

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a busy city square where hundreds of people are talking on their phones. Normally, to find out exactly where everyone is standing, you'd need a giant, high-tech microphone array or a police officer with a walkie-talkie talking to each person. But what if you could just fly a drone over the square, listen to the background chatter, and instantly figure out where every single person is without them ever knowing you were there?

That is exactly what this paper is about. It introduces a smart, autonomous drone that acts like a "silent observer" to locate multiple people using their 5G phone signals.

Here is the breakdown of how it works, using some everyday analogies:

1. The Problem: The "Blind" Drone

Usually, when drones help with 5G, they act like flying cell towers. They talk to your phone, give you internet, and then try to guess where you are. But this has a big flaw: if the drone is busy giving you internet, it can't focus on finding you. Plus, if the ground network is broken (like after an earthquake), the drone can't talk to the main network to get the data it needs.

The Solution: This paper proposes a drone that doesn't talk at all. It's a passive listener. It flies around, listens to the signals your phone sends to the cell tower, and uses those signals to find you. It's like a bat using echolocation, but instead of shouting, it just listens to the sounds others are making.

2. The Secret Weapon: The "SRS" (The Phone's ID Card)

Your phone sends out a special signal called a Sounding Reference Signal (SRS). Think of this as your phone waving a tiny, invisible flag every few milliseconds to tell the cell tower, "I'm here, and my connection is good."

The Challenge: The drone doesn't know when you wave your flag or which specific flag you are using. It's like trying to find a specific person in a crowd where everyone is waving a flag at the same time, but the drone doesn't have a list of who is waving what.
The Trick: The researchers figured out a way for the drone to recognize these flags automatically. Because the 5G standard makes these flags have a very specific, repeating pattern (like a rhythmic drumbeat), the drone can listen for that rhythm to know, "Aha! A signal is here!" even without a schedule.

3. The Magic of "Cyclic Shifts" (The Unique Voice)

Since many people are waving flags at once, how does the drone tell them apart?

The Analogy: Imagine everyone in the crowd is singing the same song. If they all sing at the exact same pitch, it's a mess. But if the conductor tells Person A to sing the song normally, Person B to sing it slightly higher, and Person C to sing it slightly lower, you can hear them all clearly.
The Tech: In 5G, this is called a Cyclic Shift. Every user gets a unique "pitch" (shift) for their signal. The drone uses a smart algorithm (called Matching Pursuit) to separate these "voices" and say, "That high-pitched song is User 1, and that low-pitched one is User 2."

4. The Hunt: How the Drone Finds the Spot

Once the drone identifies a user, it needs to find their exact location.

The Analogy: Imagine you are in a dark room trying to find a firefly. You can't see it, but you can hear a faint buzzing. You fly your drone around the room.
- When you are far away, the buzz is quiet.
- As you get closer, the buzz gets louder.
- The drone flies in a specific pattern (first a wide circle, then a tight hexagon) to map out where the sound is loudest.
The Math: The drone uses a method called Weighted Mean-Shift. Think of this as a magnet. Every time the drone hears a signal, it pulls a "virtual magnet" slightly closer to that spot. After flying around and collecting hundreds of these "pulls," the magnet settles right on top of the person's location.

5. The Results: Real-World Success

The team didn't just do this on a computer; they built a real drone and flew it in a field.

The Setup: They had 6 "phones" (User Equipments) hidden in a car and on chairs.
The Outcome: The drone flew up, listened, and found everyone.
- In open fields (rural), it was accurate to within 3 meters (about the length of a car).
- In cities with buildings (urban), it was accurate to within 8 meters.
Why it matters: This is a huge deal for emergencies. If a building collapses or a forest fire cuts off cell towers, this drone can fly in, listen to the phones of trapped people, and tell rescuers exactly where they are, all without needing to set up new equipment or ask the phones to do anything special.

Summary

This paper presents a 5G "Sherlock Holmes" drone. It doesn't need to talk to the phones or the network. It just flies around, listens to the unique "whispers" (SRS signals) that phones send out, separates them using their unique "pitches" (cyclic shifts), and triangulates their location by flying closer until the signal gets loud. It's a low-cost, high-tech way to find people when they need help the most.

1. Problem Statement

The paper addresses the challenge of localizing multiple ground users (User Equipments or UEs) in Fifth Generation (5G) networks using Unmanned Aerial Vehicles (UAVs) within the context of the Low-Altitude Economy (LAE).

Limitations of Conventional Approaches: Traditional UAV-assisted localization relies on "serving" UAVs (acting as aerial base stations). These systems often prioritize network coverage over geometric conditions required for high-accuracy positioning. Furthermore, a single serving UAV struggles to maintain Line-of-Sight (LoS) with multiple spatially distributed users simultaneously due to occlusions and flight constraints.
The Non-Serving Gap: While "non-serving" UAVs (passive receivers) offer a solution by decoupling sensing from communication, existing research often relies on idealized assumptions. Most prior works assume the receiver has "oracle knowledge" of the network's dynamic resource allocation (knowing exactly when and where users transmit). In reality, a passive UAV lacks access to the network's control plane and scheduling decisions, making signal detection, user identification, and synchronization extremely difficult, especially when multiple users transmit simultaneously in the same frequency band.

2. Methodology

The authors propose a fully autonomous framework where a single non-serving UAV captures 5G Uplink (UL) Sounding Reference Signals (SRS) to identify and localize multiple users without modifying the network infrastructure or requiring real-time control-plane access.

A. System Architecture

Hardware: The UAV is equipped with a lightweight onboard processing unit (NVIDIA Jetson Orin NX), a dual-receiver Software Defined Radio (ADALM-Pluto), and a flight controller (Pixhawk).
Operation: The UAV operates as a passive receiver. It receives minimal initial configuration (SRS parameters like periodicity and subband allocation) from the network but performs all signal processing, synchronization, and mission planning onboard.
Workflow:
1. Coarse Synchronization: Exploits the inherent time-repetition structure of 5G SRS (due to specific subcarrier spacing $K_{TC}=2$ ) to detect signal presence and align the receiver clock without prior timing knowledge.
2. Fine Synchronization & Identification: Uses Matching Pursuit algorithms on the frequency domain (DFT) to separate overlapping signals. It leverages the cyclic shifts assigned to different users in 5G to uniquely identify each UE and estimate residual timing errors.
3. Localization: Implements a Weighted Mean-Shift (WMS) algorithm. Instead of relying on Time Difference of Arrival (TDoA) or Angle of Arrival (AoA) which require wide bandwidths or complex arrays, this method uses the spatial evolution of a signal consistency metric (derived from SRS correlation) along the UAV's trajectory to estimate user positions.
4. Autonomous Mission Planning: The UAV executes a two-phase flight: an initial perimeter scan for coarse estimation, followed by a hexagonal refinement pattern around estimated locations to improve accuracy.

B. Key Technical Innovations

Blind Synchronization: The system does not need to know the exact transmission slot timing. It uses the correlation between the two identical halves of the SRS symbol to achieve synchronization.
Multi-User Separation: By utilizing the orthogonal cyclic shifts of SRS, the system can separate multiple users transmitting in the same narrowband (1.4 MHz) without interference, provided their power levels and delay spreads are within manageable limits.
Narrowband Efficiency: Unlike TDoA/AoA methods that require wide bandwidths (e.g., 40 MHz) for resolution, this method achieves high accuracy using narrowband SRS (1.4 MHz), significantly reducing hardware complexity and power consumption.

3. Key Contributions

First Practical Demonstration: The paper presents the first demonstration of practical multi-user identification and localization from real 5G UL signals captured by a non-serving UAV, operating without oracle knowledge of scheduling.
Novel Signal Processing Pipeline: Introduction of a synchronization and identification pipeline combining coarse/fine timing synchronization with a Matching Pursuit frequency extraction method to separate users in the same band.
Weighted Mean-Shift Localization: Development of a UAV-centric localization algorithm that achieves high accuracy using narrowband SRS and lightweight hardware, outperforming state-of-the-art TDoA and AoA methods by approximately 5–6 meters in simulations.
Full Autonomous Integration: Design and validation of an onboard platform capable of executing the entire sensing workflow (signal processing, identification, position estimation, and waypoint updates) autonomously during flight.

4. Results

The framework was validated through extensive simulations and a full-scale outdoor experimental campaign in both urban and rural environments.

Simulation Results (Urban):
- Achieved localization errors below 8 meters.
- Demonstrated robustness against delay and power spreads between users.
- Showed a 5–6 meter improvement in Average Localization Error (ALE) compared to benchmark TDoA and AoA methods.
Experimental Results (Rural Field Test):
- Achieved localization errors below 3 meters for users with clear Line-of-Sight (LoS).
- Maintained errors below 10 meters for users inside a vehicle (Non-Line-of-Sight/NLoS conditions), proving robustness in challenging propagation environments.
- Successfully identified and localized six simultaneous users (three in a car, three outdoors) using only 1.4 MHz bandwidth per subband.
- Mission duration averaged 2.15 minutes for initial estimates and 1.30 minutes per user for refinement.

5. Significance

This work represents a significant step forward for the Low-Altitude Economy (LAE) and emergency response operations:

Infrastructure Independence: It proves that accurate localization is possible without deploying aerial base stations or modifying existing 5G infrastructure, making it ideal for disaster zones where terrestrial networks are damaged or unavailable.
Scalability and Cost-Effectiveness: By using narrowband signals and lightweight hardware, the system is scalable and cost-effective, avoiding the high power and complexity requirements of wideband TDoA systems.
Situational Awareness: The ability to autonomously detect, identify, and locate multiple users in real-time provides critical situational awareness for search-and-rescue (SAR) missions, drone logistics, and precision agriculture.
Feasibility of Passive Sensing: The study validates that passive sensing UAVs can overcome the "synchronization gap" in cellular networks, enabling them to act as cooperative receivers for next-generation location-based services.