GUIDES · FLEET OPERATIONS

Battery health from flight logs alone.

The lithium pack is the component most likely to end a flight early and the one most fleets retire on a guess. This guide walks through Vozwin’s peer-reviewed research on predicting pack health from the logs a flight controller already keeps, and what it changes about when a pack leaves the rotation.

Last updated: September 23, 2026

Battery state of health (SoH) is the share of its original capacity a battery can still deliver, and a drone pack is considered end-of-life at 80 percent. Vozwin’s peer-reviewed research with McGill University and Université de Sherbrooke shows that SoH can be predicted within 2.26 percent using only the voltage, current, and throttle a flight controller already logs. No extra sensors, a few hundred flights, and a method that turns each flight into an image so a pretrained vision model can read it.

For a fleet, that number decides three things: how much endurance margin a pack really has for the mission it is assigned to, when it should leave the rotation, and whether the spares budget is going to packs that still had life in them. This guide explains what the research set out to solve, how the data was collected by flying rather than on a bench, why the flight logs were turned into pictures, what transfer learning contributed, what the results change on the flight line, and where the paper’s own limits lie. The paper itself is linked under Sources.

What state of health means on the flight line

Every charge and discharge cycle degrades a lithium battery a little. State of health tracks that degradation as remaining capacity: a 2,200 mAh pack that can now hold 1,980 mAh is at 90 percent SoH. Industry practice, and the paper, treat 80 percent as end-of-life, the point at which a pack should leave the rotation before a cell fails outright. Internal resistance is the other health indicator, but neither it nor remaining capacity can be measured directly in flight.

On the flight line that makes SoH the number everything else depends on. A pack at 85 percent has lost a sixth of the endurance the mission was planned against, and if the planning assumed a fresh pack, the reserve the pilot thinks they have is not there. Retire packs early and the spares budget pays for capacity never used. Retire them late, or on a cycle count that treats every pack the same, and the failure mode is an aircraft coming down mid-mission. Predict SoH accurately and both decisions become scheduling decisions with evidence behind them.

Why drone batteries are harder to model than EV batteries

Machine-learning SoH prediction is a mature topic for electric vehicles. UAVs had been largely overlooked, and not for lack of interest. The constraints are different in five ways:

Data scarcity. Training data comes from run-to-failure life-cycle tests, and for drone packs almost none had been published. Each pack has to be flown repeatedly until it dies.
Variety. A single airframe accepts packs of many brands, sizes, capacities, and voltages, and an operator typically owns several. One model has to cope with all of them.
No room for sensors. Temperature is one of the most informative signals of battery health, but a temperature sensor is extra weight, so the paper deliberately works without it. The inputs are limited to what the flight controller already records.
Human and environmental noise. Pilot habits, payload, wind, and weather change how a pack is loaded from one flight to the next, in ways an EV drivetrain never experiences.
Expensive ground truth. Remaining capacity can only be measured by fully discharging the pack, which itself damages it, so a label cannot be collected after every flight.

Large time-series models such as transformers, the state of the art for battery forecasting when data is plentiful, overfit badly on datasets this small. The research needed a method that could learn from a few hundred flights rather than a few hundred thousand.

How the data was collected: 631 flight tests

Rather than cycling packs on a bench, the team discharged them the way an operator would: by flying. A lightweight quadcopter built around a Holybro Pixhawk 4 Mini flight controller, four 2300 kV motors, and a 4-in-1 electronic speed controller flew take-off, steady flight, and landing profiles while the controller logged output voltage and current at 10 Hz. Flights spanned take-off weights of 150 to 250 grams and durations of 5 to 15 minutes to cover a realistic spread of missions. After each flight the pack was recharged on a balance charger using the same protocol every time.

Battery	Brand	Capacity	Cells	Discharge rate	Packs tested	Flight experiments
Type 1	SIGP	1,100 mAh	3S, 11.1 V	20C	5	289
Type 2	RoaringTop	2,200 mAh	3S, 11.1 V	25C	3	342

Both are lithium polymer packs typical of lightweight UAVs. The larger 2,200 mAh packs produced more flights each because they take more cycles to reach end-of-life.

Because a full discharge damages the pack, remaining capacity was measured on a battery tester only after every tenth flight, with the ten flights in between flown at the same weight and duration. The capacity for each intermediate flight was interpolated linearly between the two surrounding measurements. The paper states that assumption openly, and it is how all 631 flights ended up with a usable label.

Turning flight logs into images

The central idea of the paper is a workaround for data scarcity: instead of training a large time-series model from scratch, reshape each flight’s sensor data into a picture and let a vision model that has already learned from 14 million photographs do the feature extraction. Each flight goes through the following steps:

Each signal (voltage, current, throttle) is a one-dimensional series of measurements over the flight. It is normalized to a 0 to 255 range, the scale of a grayscale pixel.
The series is reshaped into a two-dimensional matrix 112 columns wide and rounded to whole numbers, giving one grayscale image per signal. Resolution is reduced, but the images still resolve voltage changes of 0.05 V and current changes of 0.0075 A.
The three grayscale images are stacked into a single three-channel image, one channel per signal, like the red, green, and blue of a photograph.
Because flights differ in length, images are padded with black to a common size and resized to 80 by 80 pixels, the input size used for ResNet-50.
A ResNet-50 pretrained on ImageNet, with its weights frozen, turns each image into a vector of 2,048 features. No battery-specific training happens at this stage.
Small trainable fully connected layers map those 2,048 features to a single number: the predicted remaining capacity after that flight.
The previous flight’s remaining capacity is fed in as an extra input (autoregression), so the model carries history forward instead of judging each flight in isolation.

Because ResNet-50 does the heavy lifting with weights nobody has to train, the number of trainable parameters is small enough to learn from a few hundred flights without overfitting. Hyperparameters were tuned by sampling 500 candidate configurations with Latin hypercube sampling and keeping the one with the lowest validation error over 2,000 epochs.

Transfer learning between two pack types

The second idea addresses variety. The target was the 1,100 mAh pack, which had the smaller dataset: 4 packs and 203 flights for training, and 1 held-out pack with 86 flights for testing. The 342 flights from the 2,200 mAh packs were used to pretrain the fully connected layers first, then the model was fine-tuned on the 1,100 mAh training data. In effect, the model learned what degradation looks like on one chemistry and size, then adapted that knowledge to another.

The effect shows up in the training curve. Right after the switch to the new battery data the loss ticked up briefly, then fell below the model trained without transfer learning. At the final epoch, training loss with transfer learning was 28 percent lower than without it. For a fleet, that is the property that matters: when a new pack type enters service, a model trained on the packs you already fly adapts to it with far fewer flights than starting over.

The results

The proposed approach was compared against two models commonly used for battery SoH prediction, a convolutional neural network trained from scratch and a transformer, each tuned with the same hyperparameter search. All three were evaluated on the same unseen test battery:

Model	Test MSE	Test error (MAPE)
CNN	0.0102	7.83%
Transformer	0.0175	9.22%
ResNet-50, no transfer learning	0.0053	3.47%
ResNet-50 with transfer learning	0.0027	2.26%

MSE is mean squared error; MAPE is mean absolute percentage error, the average size of the prediction miss relative to the true remaining capacity. Test set: one 1,100 mAh battery and 86 flights the model never saw during training.

Two things stand out. The transformer, the most sophisticated model in the comparison, did worst on the test set and overfit the training data heavily, which is exactly the small-data failure the paper set out to avoid. And the pretrained vision model alone already beat both baselines by a wide margin; transfer learning then cut its remaining error by a further third.

Since publication, training has continued on a larger dataset, and the test error on the current model stands at 1.57 percent.

What this changes on the flight line

Retirement becomes a data decision. Instead of pulling packs at a conservative cycle count, you retire them when the predicted capacity approaches 80 percent, and not before. That is where the spares savings come from, and it is a decision you can document in the maintenance record with the evidence behind it.
Rotation follows health, not age. Two packs bought on the same day do not degrade at the same rate. A per-pack SoH lets the healthiest packs carry the long missions and the marginal ones fly short local work until they are retired, instead of every pack being treated as the fleet average.
Endurance margins are real again. Mission planning against the predicted capacity of the pack actually on the aircraft, rather than its nameplate, restores the reserve the pilot is counting on. A 2 percent prediction error is a margin you can plan around; a cycle count is not.
It works on data you already have. Voltage, current, and throttle are logged by any modern flight controller. There is no hardware to add and no weight penalty, so any solution that asks you to instrument your aircraft first should explain why.
New pack types are cheap to add. Transfer learning means a model trained on the packs you already fly adapts to a new type with a fraction of the flights, so a fleet is not locked to one brand or one capacity.
Accuracy should be reported on unseen batteries. The 2.26 percent figure is on a pack the model was never trained on, which is the only number that predicts real-world performance. A headline accuracy measured on the training packs is not comparable.

Limits and what comes next

The paper is careful about its scope, and anyone reading it should be too:

Only discharge cycles were modelled. Charging data is rarely available from UAV batteries in practice, so it was not collected.
One airframe, two pack types. Broader validation across drones and chemistries is the obvious next step, and it is what the continued training since publication has been doing.
Labels between full-discharge checks are interpolated, which assumes degradation is roughly even across ten similar flights.
Temperature, one of the strongest health signals, was left out on purpose to respect the weight constraint. Where a platform does log it, it should help.

The motor is the next component. Vozwin’s follow-on work predicts motor state of health from the accelerometers, gyroscopes, barometer, and GPS a drone already carries, plus camera footage to catch the visual instability a degrading motor produces. The fleet-level picture, including what the regulations require of your records and what to demand from any solution before it touches your aircraft, is in the companion guide on predictive maintenance for UAV fleets.

Who did the research, and where Vermilion fits

The paper was authored by Jiarui Xie, Lingchen Kong, and Yaoyao Fiona Zhao of McGill University’s Department of Mechanical Engineering; Mohamed Rami Latreche and Elaine Mosconi of Université de Sherbrooke’s Department of Information Systems and Quantitative Management Methods; and Sean Smith of Vozwin, the founder of Vermilion. It is part of the PHUMS project established by Vozwin, with UAV design and experiments guided by Fabio Bandera of Vozwin, and it was supported by the MITACS Accelerate program and the Réseau SDG Innovation Network’s collaborative R&D projects in digital, intelligent, and sustainable transformation.

Vozwin owns the technology presented in the research, which was carried out through its Labs division. Vermilion is the production platform built on it: a physics-informed reasoning model, incubated in Vozwin’s venture studio, now operating as its own company, and available for deployment on edge, cloud, or air-gapped infrastructure. In Vermilion, a data-driven estimate like this one is delivered with the reasoning and confidence a maintenance lead needs to act on it; the platform is described on the Vermilion page of this site. Vozwin Aerospace works the aircraft side: the UAV platforms we build log these signals from the first flight, and for operators with a mixed fleet we can scope how the packs you fly, the records you keep, and the model fit together.

Frequently asked questions

What is battery state of health (SoH)?

The fraction of its original capacity a battery can still deliver, so a 2,200 mAh pack that now holds 1,980 mAh is at 90 percent SoH. Drone batteries are generally considered end-of-life at 80 percent, before a cell fails outright.

When should a drone battery be retired?

Industry practice and the paper treat 80 percent state of health as end-of-life. Most fleets approximate that with a cycle count, which retires healthy packs early and tired ones late because packs do not degrade at the same rate. A per-pack prediction lets you retire each one when its predicted capacity approaches 80 percent, and record the evidence in the maintenance log.

How accurately can machine learning predict a drone battery’s state of health?

Vozwin’s peer-reviewed research with McGill University and Université de Sherbrooke reached a 2.26 percent mean absolute percentage error on a battery the model had never seen, from a dataset of 631 flights. Continued training on more data has since brought that to 1.57 percent.

What data does the model need?

Only voltage, current, and throttle from the discharge cycle, recorded by the flight controller at 10 Hz during normal flights. No temperature sensor or other added hardware is required, which matters because every extra sensor costs a drone weight and flight time.

Why convert flight data into images?

Because large time-series models overfit when data is scarce. Reshaping each flight’s signals into an image lets a ResNet-50 pretrained on 14 million photographs extract features with no battery-specific training, leaving only a small set of layers to train on a few hundred flights.

What is transfer learning in this context?

Pretraining the model on one battery type (342 flights of 2,200 mAh packs), then fine-tuning it on a second type with less data (1,100 mAh packs). It cut training loss by 28 percent and test error from 3.47 to 2.26 percent, and it means a new pack type can be added to a fleet’s model with far fewer flights.

Where can I read the full paper?

It is published on arXiv under the title Machine Learning-Based Battery State-of-health Prediction for Unmanned Aerial Vehicles Predictive Maintenance (arXiv:2607.06791), linked under Sources on this page.

Sources

↑ TO THE TOP

Retiring packs on a cycle count?

Tell us which packs you fly, how you decide when one leaves the rotation, and what your aircraft log. We will tell you what a per-pack forecast would change, and whether Vermilion or a Vozwin Aerospace integration is the right starting point.