ECE capstone design project: drone detection system

Posted on April 01, 2017


Here's an interesting example of an ECE capstone design project. It's a system for detecting the presence of drones by identifying the electromagnetic footprints of their RF control systems. Pretty clever.

  • Video transcript:

    So, can you tell us a little bit about your project?

    There's kind of like a problem in terms of detecting drones, or calculating drones, in airspace where there should be no drones. So, our project is more or less focussed on figuring out a way to detect drones at a certain frequency.

    The way it works is there should be a probability of how certain your program is that there is a drone in the area.

    With our program, it's been consistently detecting a drone.

    So, can you show me some of the technical aspects of your design?

    Yeah, so we have a controller, a drone crontroller. Basically, what it does is connects with our drone but also it's emitting a signal in a low frequency band. So, in our case it's 27.145 MHz and it's being captured by our software-defined radio so in this case we are using a HackRF radio and then we perform signal analysis on the data that's being collected.

    So after you collect all the data, what are the next steps in terms of analyzing it?

    They way we're detecting a drone presence is by trying to find the signals the drones use to communicate and to capture them and see if it really looks like a drone

    So, what the software-defined radio does is it brings in the time data, so if we capture a second worth of data we have a sample rate. So, in a second we can pick up two million data points. Which are picking up the frequencies.

    But the time data doesn't tell us all the information that we need to detect the drone. So we compute a Fourier analysis on the time data to convert the data from the time domain to the frequency domain. When the data is in the frequency domain we can see the power level of signals operating at different frequencies. That's called the frequency spectrum.

    When you look at the frequency spectrum, you can see which frequencies have the most prominent signal levels. So, if there's a drone operating at 27.145 MHz and you look at the frequency spectrum from 27 MHz to 28 MHz you'll see a big spike at the 27.145 MHz coordinate. But a spike doesn't always mean a drone.

    So, what we do, we take our input data, grab the time data, convert it to the frequency domain and then we take the input frequency domain and we compare it to what we know a drone shape looks like already in the frequency domain. One we have our input data, we scan across and we see if there's any section of the frequency spectrum that looks like a drone and then it gives us a really easy way to compute the probability of a drone.

    So, can you tell me a little bit about the software that actually runs your project?

    Yeah, sure.

    So, our industry partner gave us essentially requirements for what our software is supposed to accomplish and so we built our software based off of those requirements the intention is that our client takes our software and essentially builds upon it.

    What would say then are the next steps for this design?

    One aspect that we can improve in our project is the drone template that we're using.

    Because the drone template that we're using right now is based off of a couple of experiments in an environment that may have interference in it.

    Another feature that we want to implement into our project is identifying that modulation scheme of a controller chipset. Some drones carry the same kind of chipsets, so we're able to identify drones by their chipsets.