Micro drone sensor/detector

[zwirek2201]

Hi guys!

TLDR: Tiny drone, 65mm, 20g, we want to detect that it flew through a ring without putting any marker/tag on the drone itself. Shouldn't trigger if the drone flies right outside of the ring.

I'm a race director at Polish national micro drone racing league and for a long time I wanted to figure out a way to detect that a drone flew through a gate. Right now our laptiming system uses 5.8GHz video signal strength from the drone to figure out when it flew next to the laptimer and it counts that as a lap. Sometimes, though, the drone can fly right next to the laptimer, but not fly through the start gate and in that case the lap should not be counted, right now it's manually verified.

The biggest problem is that we don't want any additional sensors/tags on the drones because pilots fight for every 0.1g in their weight and these rockets are both small (about 65mm) and fast (easily doing 50kph indoors).

I tried using laser Time of Flight sensors with some success, but they usually have a really narrow FOV so you need multiple to cover the gate. They have to be calibrated for gate size/shape so they don't trigger when the drone flies right next to the gate, and they work about 80% of the time, if the drone is going REALLY fast, it sometimes won't detect it.

I've heard of some people experimenting with other sensors, but never with success. I was wondering if it would be possible to embed some loop in the gate and use that for a detection mechanism, but my physics/electronics knowledge is somewhat limited.

Video from one of our pilots for reference: 

Do you have any ideas how to tackle this?

[ki8aras]

i mean im not very into this type of stuff, but it sounds to me you just need a laser with a detector on the other side, and record when contact is interrupted no?

[zwirek2201]

3 minutes ago, ki8aras said:

i mean im not very into this type of stuff, but it sounds to me you just need a laser with a detector on the other side, and record when contact is interrupted no?

Yeah, that's what I started with, kind of how I used the time of flight sensor, but to cover the entire gate we'd need a whole array of emitters and detectors on the other side of the gate. In case of circular gates, that's pretty difficult to do. Usually for that I would try to use infrared, but LEDs in gates sometimes mess with IR and can give a lot of IR noise. 

[FlyingPotato_is_taken]

Few options you could consider:

1. light curtain

2. Metal detector.

3. High speed global shutter Camera with image recognition. This could be the easiest option as you already have an video feed for manual analyse (right?). So all you need is pipe this stream through some software.

4. 5.8Ghz triangulation. Probably the most difficult solution and could get you into the dual-use restrictions/paper work. If you could modify the existing 5.8Ghz tracker to an external trigger all you would need is a small FPGA that generates the three trigger signals and handle the output to triangulate the position of the drone. Issue here is the real time aspect of this solution. There can't be significant delay between the three sensors and normal SDR aren't optimized for this application.

5. Microphone/sound triangulation. You might be able to buy a commercial solution as this as far as I know how it is done on shooting ranges.

6. Radar. Not sure if the none military/dual use sensor could catch the drone. They are quick and small.

 

6 hours ago, zwirek2201 said:

Yeah, that's what I started with, kind of how I used the time of flight sensor, but to cover the entire gate we'd need a whole array of emitters and detectors on the other side of the gate.

ST has time of flight sensors. 

VL53L0X and Arduino should be enough to test if these sensors could detect a drone. I doubt they can do this as they are primarily used for gesture recognition and some range measurements.

 

 

To be honest: Talk to other race directors. Somebody has probably already figured out how this could be done economically.

[zwirek2201]

3 hours ago, FlyingPotato_is_taken said:

Few options you could consider:

1. light curtain

2. Metal detector.

3. High speed global shutter Camera with image recognition. This could be the easiest option as you already have an video feed for manual analyse (right?). So all you need is pipe this stream through some software.

4. 5.8Ghz triangulation. Probably the most difficult solution and could get you into the dual-use restrictions/paper work. If you could modify the existing 5.8Ghz tracker to an external trigger all you would need is a small FPGA that generates the three trigger signals and handle the output to triangulate the position of the drone. Issue here is the real time aspect of this solution. There can't be significant delay between the three sensors and normal SDR aren't optimized for this application.

5. Microphone/sound triangulation. You might be able to buy a commercial solution as this as far as I know how it is done on shooting ranges.

6. Radar. Not sure if the none military/dual use sensor could catch the drone. They are quick and small.

 

ST has time of flight sensors. 

VL53L0X and Arduino should be enough to test if these sensors could detect a drone. I doubt they can do this as they are primarily used for gesture recognition and some range measurements.

 

 

To be honest: Talk to other race directors. Somebody has probably already figured out how this could be done economically.

Thanks for the answer, some really good ideas.
 

The unfortunate thing is that no one really figured it out. Some guys in UK have some technology (I'm pretty sure it's metal detection based) with PoC, but it's still not good enough for actual race use and they've been working on it for over a year. It's a pretty young hobby and there are still only a few larger leagues around the world.

VL53LOX is actually the sensor I used. I would need 3 of them to cover the entire gate, it's definitely possible and I had PoC working a couple of months ago, but it was not reliable on higher speeds. You can increase polling rate, but it's still not quite there. 

5.8GHz triangulation has been my dream for some time as it would allow us to do some more advanced tracking and maybe pipe the data to make a map for our streams. It could maybe be possible, but it's definitely out of my knowledge level right now. I think it would be tricky because transmitters on drones are constantly tweaking their output level, so that would require some additional processing.

Image processing is probably the route that I'm going to investigate next. Either external hi speed camera or processing the FPV video that comes from the drone. Probably a decently simple algorithm could recognize a gate and figure out if the drone flew through it. It would have to process 4 video streams at the same time, but with some optimization and decent hardware it should be possible.

Thanks!