Could The Electric Motors In Drones Give Away Their Location?

Researchers have developed a system that can detect small drones by listening to the magnetic fields generated by their motors. The technology is inexpensive and doesn't require any transmitted signals, but there are still questions about how useful it could become.

Researchers at Florida International University (FIU) have developed a new way of detecting small drones, using the magnetic fields produced by the electric motors that drive their propellers. The system uses a sensitive receiving antenna and machine learning to recognise different aircraft, and has achieved detection accuracy of up to 98.52% in outdoor testing.

The research, highlighted by FIU on October 23, was carried out using three commercially available micro-drones, with the equipment detecting their presence at distances of up to 15 metres. The researchers say their prototype costs only a few hundred dollars to build, and they are now working with industry to develop the technology further.

There are already several methods of detecting drones, including radar, acoustic sensors and cameras. Radio-frequency monitoring is another option, although that depends on the aircraft transmitting a signal that can be recognised. The FIU system is a little different because it detects electrical activity associated with the motors themselves.

It's an interesting approach, particularly as electromagnetic interference is something that electronics engineers normally spend considerable effort trying to reduce. In this case, the researchers have found a way of using those unwanted emissions to identify the equipment producing them.

How the researchers are detecting drones

Most small multirotor drones use brushless DC motors, with an electronic speed controller regulating the current supplied to each motor. By switching current through different windings, the controller produces a rotating magnetic field that drives the rotor and its attached propeller.

Of course, the magnetic field doesn't remain entirely inside the motor. Some of it escapes into the surrounding environment, along with emissions associated with the current flowing through the motor windings and electrical connections.

For an electronics engineer, this is hardly a surprising discovery. Motors, power supplies and switching circuits can all produce electromagnetic interference, and there are numerous design techniques intended to limit its effects on other equipment.

However, there's information buried in that interference. The electrical activity inside a motor isn't random, and the frequencies present in its emissions can be influenced by its construction, speed and the way it is driven.

This is what the FIU researchers have been investigating. Different drones use different combinations of motors and controllers, which means their electrical emissions may contain patterns that can be used to tell them apart.

The system uses a circular loop antenna to pick up extremely low-frequency magnetic signals. A loop antenna operates through electromagnetic induction, with a changing magnetic field generating a voltage in the conductor. The signals received from small drone motors can be weak, particularly when the aircraft is some distance away, so the researchers have also developed a low-noise analogue front end to condition them.

The next part involves identifying useful patterns in the received signal. Instead of measuring signal strength at a particular frequency and deciding whether a drone is present, the researchers use spectral correlation analysis to look for repeating features associated with the operation of the motors.

This is particularly useful for machinery that operates in a periodic manner. The switching and rotation of an electric motor can produce related frequency components, even when those signals are mixed with other electrical noise.

The resulting data is processed using a convolutional neural network based on VGG16, which has been trained to identify the characteristic patterns of different drones.

In the study published in IEEE Sensors Journal, the researchers reported outdoor detection accuracy of up to 98.52%, although identifying which drone was present proved more difficult.

When the system was asked to distinguish between three drone types and the absence of a drone, it achieved an overall accuracy of 83.5%. The researchers also found that their smallest test aircraft was harder to recognise because it produced weaker magnetic emissions.

That's an important limitation. Identifying the presence of electrical activity from a drone is one thing, but recognising a particular aircraft from its motor emissions becomes much harder when the signal is weak or resembles something else in the environment.

Nevertheless, the results suggest that there may be useful information available from electrical emissions that would ordinarily be regarded as unwanted noise.

Another advantage is that the system doesn't need to transmit anything. Radar operates by sending out electromagnetic signals and measuring their reflections, while the FIU equipment only receives emissions already being generated by the drone.

It also doesn't depend on intercepting a drone's radio control system. This could be useful when dealing with autonomous aircraft or drones using alternative communications methods, provided their motors generate a sufficiently strong and recognisable magnetic signal.

Could magnetic sensing become a practical way to detect drones?

There is an obvious problem with the research in its present form. A maximum demonstrated distance of 15 metres provides very little warning of an approaching aircraft, particularly when dealing with a fast-moving drone.

For comparison, conventional counter-drone radar systems can detect aircraft at distances far beyond those demonstrated by the FIU researchers. Even relatively inexpensive cameras and acoustic sensors may provide useful information at greater distances, depending on the environment.

So while the ability to identify a drone from its magnetic emissions is impressive, the technology isn't yet a replacement for established detection systems.

However, there may be situations where short-range detection is still useful.

Consider a facility that needs to monitor several possible entry points or locations where an aircraft might approach at low altitude. Installing radar equipment to cover every position could become expensive, particularly if the site is large or contains buildings that obstruct the available coverage.

A network of inexpensive magnetic sensors could potentially provide an additional layer of detection in places where other equipment has difficulty identifying small aircraft.

There are some interesting possibilities for protecting infrastructure, including power stations, industrial facilities and other locations where unauthorised drones might present a problem. The sensors would need to be positioned relatively close to the areas being monitored, but their low cost could make deploying multiple units more practical.

Of course, placing magnetic sensors around industrial facilities introduces another challenge.

An industrial environment is hardly free from electromagnetic interference. Electric motors, transformers, power converters and other equipment all generate changing magnetic fields. A sensor installed near a large motor drive might receive a far stronger signal from the machinery than from a small drone passing nearby.

The researchers have demonstrated that their signal-processing approach can distinguish drone emissions from background measurements under the conditions they tested. Whether it can maintain that performance alongside numerous other electrical devices remains an important question.

There is also the problem of detecting unfamiliar aircraft.

Machine-learning systems can become very good at identifying patterns included in their training data, but recognising a completely different drone is another matter. Changes in motor speed, flight manoeuvres and electrical loading could also influence the emissions received by the antenna.

For a practical security system, detecting the presence of an unknown drone might ultimately be more important than identifying its exact model. A system that only reliably recognises aircraft it has previously encountered would have limited value.

The short detection range is probably the hardest problem to solve. Magnetic fields from compact sources generally become much weaker as distance increases, with the exact behaviour depending on the source geometry and operating conditions.

Increasing receiver sensitivity might help, although that also makes the system more vulnerable to unwanted signals. Larger antennas, better filtering and more sophisticated processing could offer improvements, but none can remove the underlying difficulty of detecting a weak magnetic signal against background noise.

Nevertheless, I think there is something particularly clever about this research.

Drone detection usually involves looking for something associated with the aircraft, such as its shape, sound or radio transmissions. The FIU researchers are instead examining the electrical behaviour of the machinery keeping it airborne.

That could become increasingly relevant as autonomous drones and aircraft using unconventional communication links become more common. A drone might operate without transmitting a conventional control signal, but an electrically powered multirotor aircraft still needs to energise its motors.

The FIU team is now working with industry to develop the technology further. Whether it eventually becomes a practical counter-drone product will depend on improving its range and demonstrating reliable operation in more demanding environments.

Even if it remains limited to close-range detection, the research offers an interesting example of how electromagnetic interference can be put to useful work. Engineers have spent decades trying to stop unwanted emissions from escaping electronic equipment. Here, those same emissions are providing a way to recognise a machine without ever making contact with it.