⚡ electrical engineering

Time-Constrained Erasure Correction for Data Recovery in UAV-Aided IoT

This paper proposes adaptive data-recovery schemes utilizing fountain coding and message replication to enhance uplink reliability in UAV-aided IoT networks by dynamically adjusting redundancy based on sensor-UAV contact times, specifically within a LoRa-based random access system activated by wake-up radios.

Original authors: Kushwanth Sistu, Siddhartha S. Borkotoky

Published 2026-01-28

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Kushwanth Sistu, Siddhartha S. Borkotoky

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

The Big Picture: The Drone and the Village

Imagine a small village (the IoT network) where every house has a sensor (like a thermometer or a moisture detector) that needs to send a report to a central office. Instead of building a massive tower to catch all the signals, the village uses a drone (the UAV) that flies over the village, hovers for a short while, and acts as a temporary mailman to collect these reports.

The problem is that the sensors are low-power and the connection is shaky. It's like trying to shout a message across a windy field; sometimes the wind (interference) or distance causes the message to get lost. If the drone flies away before it hears everything, that data is gone forever.

The Solution: Sending Extra Copies

The authors propose a smart way to make sure the drone gets all the data, even if some messages get lost. They use two main tricks, both based on the idea of redundancy (sending extra stuff just in case).

Think of it like sending a letter through a chaotic post office where letters often get lost.

The "Copycat" Method (Message Replication): You write your letter three times and put them in three different envelopes. If the post office loses one or two, the third one might still arrive.
The "Puzzle" Method (Fountain Coding): Instead of sending copies of the same letter, you cut your letter into pieces, mix them up, and send out many different "scrambled" envelopes. As long as the post office delivers enough of these scrambled pieces (even if they are different ones), the receiver can mathematically put the puzzle back together to read the original letter.

The Smart Twist: Adapting to Time

The most important part of this paper is that these methods are adaptive.

Imagine the drone is hovering over the village for a specific amount of time (say, 30 seconds).

If the drone stays a long time: The sensors know they have plenty of time. They can afford to send lots of extra copies or puzzle pieces to be super safe.
If the drone is about to leave: The sensors realize, "Oh no, we only have 5 seconds left!" They immediately stop sending extra copies because there isn't enough time to send them, and sending too many might just cause a traffic jam (collisions) where messages crash into each other and get lost anyway.

The system constantly checks the clock and decides: "Do we have time to send a backup? Yes? Send it. No? Just send the main message and hope for the best."

The Setup: Waking Up the Sensors

To save battery, the sensors in the village are usually asleep. They have a tiny, super-low-power "doorbell" (a Wake-Up Radio).

The drone arrives and rings the doorbell (sends a Wake-Up Call).
The sensors hear the bell, wake up their main radio, and start shouting their messages.
The drone listens, collects the data, and flies away.

What the Experiments Showed

The authors ran computer simulations to see how well this works compared to other methods.

Better than "Random Shouting": Without these tricks, sensors just shout randomly. If two sensors shout at the same time, their voices cancel out (a collision). The new methods fix this by sending backups intelligently.
The Puzzle vs. The Copy: The "Puzzle" method (Fountain Coding) generally worked better than the "Copycat" method (Replication). Why? Because with puzzles, every single piece you receive helps you solve the whole picture. With copies, if you get three identical copies of the same letter, it doesn't help you if the original was lost; you just need one copy.
The Trade-off:
- If the drone hovers for a short time, simple copying works okay, but complex puzzles might fail if too many pieces are lost.
- If the drone hovers for a long time, the puzzle method shines because it can recover from many lost pieces.
- If there are too many sensors crowded together, everything gets messy (collisions), but the puzzle method still handles the chaos better than the others.

The Bottom Line

This paper presents a smart, flexible way for drones to collect data from ground sensors. Instead of using a rigid, one-size-fits-all approach, the sensors adjust their strategy based on how much time the drone gives them. This ensures that even in a noisy, crowded, and short-lived connection, the most important data gets delivered.

Technical Summary: Time-Constrained Erasure Correction for Data Recovery in UAV-Aided IoT

Problem Statement
The paper addresses the challenge of improving uplink reliability in Internet of Things (IoT) networks where Unmanned Aerial Vehicles (UAVs) act as mobile gateways collecting data from ground-based sensor nodes. In typical UAV-aided scenarios, sensors transmit data to a hovering UAV, which then relays it to a sink. However, these sensor-to-UAV links are inherently lossy due to low transmission power, interference, and the limited duration of the UAV's hovering window.

Existing erasure-correction methods, such as message replication and fountain coding, face specific constraints in this environment:

Time Constraints: The UAV's limited hovering time restricts the number of redundant frames a sensor can transmit.
Collision Risks: In a multi-sensor cluster using random access, excessive redundancy can increase frame collisions, potentially degrading performance rather than improving it.
Variable Conditions: Sensors may wake up at different times during the UAV's visit, leading to variable remaining contact times.
Protocol Limitations: While Time Division Multiple Access (TDMA) can eliminate collisions, it requires a reliable connection setup phase (learning sensor presence and data size), which adds complexity and latency. Random access avoids this but suffers from collisions.

Methodology
The authors propose two adaptive erasure-correction schemes designed for a random-access scenario using Wake-up Radios (WuR) and Long Range (LoRa) technology. The system model assumes $n$ sensors, each with a LoRa transceiver and a low-power WuR. The UAV hovers for $N_s$ time slots, broadcasting Wake-up Calls (WuC) containing slot information. Sensors wake up upon receiving a WuC and must transmit $\beta$ messages within the remaining slots.

The core innovation is the adaptive selection of redundancy based on the remaining contact time ( $\gamma(i)$ ) available to a sensor that wakes up in slot $i$ . The schemes transmit up to $\beta + \varepsilon$ frames, where $\varepsilon$ is the redundancy budget.

Fountain Coding Scheme:
- If the remaining slots $\gamma(i)$ are sufficient to transmit $\varepsilon$ redundant frames ( $\gamma(i) \ge \varepsilon$ ), the sensor encodes the $\beta$ source messages into $\beta + \varepsilon$ linear combinations (coded messages) using random linear fountain codes over a Galois field. These are transmitted over randomly chosen slots.
- If $\gamma(i) < \varepsilon$ , fountain coding is skipped. The sensor transmits as many uncoded messages as possible within the remaining slots.
- Decoding succeeds if the receiver obtains $\beta$ linearly independent combinations.
Message Replication Scheme:
- If $\gamma(i) \ge 0$ , the sensor calculates the number of full copies ( $m_q$ ) and the remainder ( $m_r$ ) of messages that can be sent. Messages are replicated $m_q+1$ or $m_q$ times depending on the available redundancy budget.
- If $\gamma(i) < 0$ (insufficient time for all original messages), the sensor randomly selects a subset of messages to transmit, discarding the rest.

Key Contributions
The paper claims the following contributions:

Novel Schemes: It devises two erasure-correction schemes specifically for UAV-IoT data recovery: one using fountain coding and one using message replication. The authors state this is the first investigation of its kind for this specific scenario.
Adaptive Redundancy: Unlike conventional static erasure correction, these schemes dynamically determine the amount of redundancy based on the sender's remaining contact time with the UAV, allowing adaptation to variable hovering durations.
Practical Design: The design is illustrated for a practical setup involving WuR for activation and LoRa for data transfer with random access.
Analytical Modeling: The authors develop analytical models to predict the message delivery probability (MDP) for both schemes, accounting for slot-dependent transmission success probabilities and collision dynamics.

Performance Results
Numerical results, obtained via MATLAB simulations with 30 sensors, a 10m UAV altitude, and Nakagami-m fading, highlight the following:

Comparison with Baselines: Both proposed schemes consistently outperform a standard random-access scheme (without erasure correction).
Fountain vs. Replication: Fountain coding generally outperforms message replication because each coded frame carries information about all source messages, whereas replication only protects specific frames. However, fountain coding is sensitive to the number of erasures; if the number of lost frames exceeds the redundancy capacity, decoding fails.
Impact of Parameters:
- Wake-up Probability ( $P_b$ ): The proposed schemes are robust even when $P_b$ is low. In contrast, a TDMA benchmark performs poorly when $P_b$ is low because sensors fail to wake up and receive time slot assignments. TDMA only outperforms the proposed schemes when $P_b > 0.9$ .
- Hovering Time: Performance improves with longer hovering times due to reduced collision probability. For short hovering times, message replication with low redundancy ( $\varepsilon=1$ ) offers minimal improvement, while fountain coding can underperform if the redundancy is insufficient to correct multiple erasures.
- Node Density: As the number of sensors increases, collision rates rise, degrading all schemes. However, with sufficient redundancy ( $\varepsilon=3$ ), the proposed schemes maintain an advantage over the baseline.

Significance and Claims
The paper concludes that these schemes offer a method to improve data-collection performance in UAV-IoT networks without incurring the high complexity overhead of TDMA or the reliability risks of pure random access. By adapting redundancy to the specific time constraints of the UAV's visit, the schemes are positioned to benefit low-altitude use cases such as precision agriculture and infrastructure monitoring. The authors modestly frame the work as a design philosophy and analytical framework that addresses the gap in existing literature regarding erasure correction under variable contact times and energy constraints in UAV-IoT clusters.