Author Topic: Brainstorming about bachelor thesis project - Data transmission using car lights  (Read 2566 times)

0 Members and 1 Guest are viewing this topic.

Offline 97hilfelTopic starter

  • Regular Contributor
  • *
  • Posts: 66
  • Country: it
Hello everybody,
I'm a computer science student that has an EE background, thus I choose to do a bachelor thesis about the viability of a data channel for cars using their lights.
I know this has already been done in some way, for example, LiFi but I am not certain if that would also work in my environment.
The basic setup is to modulate the light of the car and pick it up with some kind of camera or sensor. The idea is that a car authenticates itself using its light to another car via its led taillights and the car behind it can pick it up with a special sensor or ideally with the camera most cars already have. This is really meant more as a proof of concept and construction of a prototype.
Thus I wanted to ask if somebody has ideas about what sensors and or camera modules I could take a look at.
I already have the raspberry pi cam in mind and I have some spare Logitech C920's and an Intel RealSense D435 around.
For the control of the led taillight I honestly just wanted to use an off the self demo-board and a decent MOSFET that can drive the thing.
 

Offline sokoloff

  • Super Contributor
  • ***
  • Posts: 1799
  • Country: us
I would think that any optical link would likely be done via IR rather than visible light for compatibility with humans coupled with high enough intensity to be detectable in daylight.
 

Offline 97hilfelTopic starter

  • Regular Contributor
  • *
  • Posts: 66
  • Country: it
The point is to try it with the already installed on a vehicle. Transmission speed is not a primary conern.
 

Online tunk

  • Super Contributor
  • ***
  • Posts: 1099
  • Country: no
I would guess that at least some taillights use PWM and full intensity for brakes.
 

Offline DaJMasta

  • Super Contributor
  • ***
  • Posts: 2393
  • Country: us
    • medpants.com
It could be tough to make universal - there's a big variety in tail lights used and driving mechanisms (not many relay drive ones left, but can you imagine the thrashing  :-DD ) .  Depending on bulb type, you could also run into switching issues if you were using particularly high intensity lights, different types will put in different limitations, so a "universal" drop in system could be pretty low performance compared to, for example, an all LED light car.


I think the bigger block could be the camera technology, actually.... because of framerate.  Unless you're going to PWM the lights to encode the data as effectively analog, your 60fps camera would struggle to get anything more than 30bps datarate.  If you could shrink the resolution to get more framerate you can probably do better, but you're still talking a few hundred frames per second, tops, so your datarate is still low enough that a unique identifier for the vehicle and overhead for a small ID packet being transmit could take multiple seconds.  There's also the issue that if your data is being transmit this slowly, it will manifest as flicker which could be distracting/unsafe on the road.  If you can get the frequency to a few hundred Hertz you can avoid it, but then your camera has to be specialized to be able to pick it up.

The typical optical data link is a set frequency pulse modulated with the data, basically the same idea as a radio link, and I don't think cameras will be able to manage that.  If you're going to try to develop a device to find such a signal, you need to make sure you can actually modulate your lights at the required speed, and you probably want to be looking at camera cores that offer 240x180 or lower resolution modes at a few hundred fps.
 

Offline sokoloff

  • Super Contributor
  • ***
  • Posts: 1799
  • Country: us
If you can insist on LED tail lights, are willing to have them visible to following drivers (in a way that's not distracting or confusing), you could try modulating them at 38kHz and use a COTS IR receiver with the visible light filter removed and see if you could establish a link that way.

(I get that this is a thesis, so some amount of practical constraints might not apply, but I think you'll find in doing the above in a variety of lighting conditions that you get better results with a high power directional IR LED modulated at 38kHz. The above is perhaps a thesis-approvable step along that journey. Won't work at all with incandescent/halogen lamps.)
 

Offline mark03

  • Frequent Contributor
  • **
  • Posts: 749
  • Country: us
There is plenty of literature in, e.g., IEEE journals on high-speed data using "ambient" lighting infrastructure, like the white LEDs providing room illumination.  I was surprised to learn that even with the phosphor, white LEDs work fine for high-speed data.  All of the practical/effective systems use pulse-position modulation (PPM).  There are even tricks to make this compatible with PWM dimming, by varying the data encoding to change the average intensity.  In office environments this could all be made to work quite well; the "problem" is that Wifi got so good, so quickly, that there just wasn't a good use case for optical, so none of the proposed LiFi systems saw the light of day (so to speak :))

A camera is the wrong tool for the receiving end, unless the data will be spatially as well as temporally encoded (hmmm, rapid-fire QR codes on the tail lights, anyone? ;)  Normally you would use a photodiode.  For simple/slow proof of concept, I think the suggestion to build off of a standard IR remote-control protocol is a good one.  You can use the same protocol even if you are modulating a visible LED, and the speeds make it relatively easy.
« Last Edit: October 24, 2019, 04:13:49 am by mark03 »
 
The following users thanked this post: I wanted a rude username

Offline Bud

  • Super Contributor
  • ***
  • Posts: 7276
  • Country: ca
Don't forget to tell your car insurance company you are modifying your car  >:D. I am sure they will be happy to ramp your imsurance premium up  ^-^
Facebook-free life and Rigol-free shack.
 

Offline 97hilfelTopic starter

  • Regular Contributor
  • *
  • Posts: 66
  • Country: it
If you can insist on LED tail lights, are willing to have them visible to following drivers (in a way that's not distracting or confusing), you could try modulating them at 38kHz and use a COTS IR receiver with the visible light filter removed and see if you could establish a link that way.
Maybe I should have been clearer, the point is to try and use the existing led lights, with a different driver circuit and try to make it invisible to following drivers, still using normal light, I know this is probably not possible but thats how the professor wants it to be tested. This will be a first lab setup and not a road setup, I guess for that they would have called it a masters thesis and not a bachlors thesis.

Don't forget to tell your car insurance company you are modifying your car  >:D. I am sure they will be happy to ramp your imsurance premium up  ^-^
As mentioned, its a lab setup, no car will be harmed ;)

A camera is the wrong tool for the receiving end, unless the data will be spatially as well as temporally encoded (hmmm, rapid-fire QR codes on the tail lights, anyone? ;)  Normally you would use a photodiode.  For simple/slow proof of concept, I think the suggestion to build off of a standard IR remote-control protocol is a good one.  You can use the same protocol even if you are modulating a visible LED, and the speeds make it relatively easy.

I'm certain that a camera is not the right tool for this job, but the professor wants a camera with a transmission rate of approximate max. 10bps. This system should be used as a tool to identify cars as 2nd factor before starting wireless communication, due to the nature of such a system such a signature is hard to forge and would represent a reliable tool, at least in good lighting conditions :)
 

Offline max_torque

  • Super Contributor
  • ***
  • Posts: 1327
  • Country: gb
    • bitdynamics
Modern cars use PWM LED lights, front and rear, and hence your data bit modulation of the light is going to have to be faster, or slower, than that existing current control modulation!

Watch any film of a modern car and see how the camera picks up its lights as flicker, or strobe-ing already!

Usually, the current control driver is already built into the individual light assembly, so it will be difficult, but not impossible, to remove that driver and modify or change it. I suggest buying some second hand led light units (from ebay for example) and reverse engineering them as a first step. You may get lucky, and be able to remove the existing control pcb, and replace it with your own version, that can encode data by changing the actually modulation frequency for example (which would be fairly simple to decode with a PLL based receiver or similar)



« Last Edit: October 24, 2019, 12:53:17 pm by max_torque »
 

Offline StillTrying

  • Super Contributor
  • ***
  • Posts: 2850
  • Country: se
  • Country: Broken Britain
about the viability of a data channel for cars using their lights.

That's easy, it's not viable. :)
.  That took much longer than I thought it would.
 

Online tunk

  • Super Contributor
  • ***
  • Posts: 1099
  • Country: no
I have no experience with this, but I would guess when you're
going to use a camera (30fps?) and 10bps that the only way is
to vary the taillight intensity. If you could make the variation
small enough (maybe less than 5%) then it may appear invisible.
« Last Edit: October 24, 2019, 02:22:37 pm by tunk »
 

Online LaserSteve

  • Super Contributor
  • ***
  • Posts: 1355
  • Country: us
Sync the camera to the trailing edge of tbe gps second epoch. Then use Gold and Barker codes as the sync preamble as the leds will be synced as well. . Basically spread spectrum modulation. Use multiple leds or red and amber or IR for the in phase and quadrature. 5G phones can also supply an inherent clock as can the carrier of AM radio stations. Use the VIN as the sysID.  Now award me my patent and my BSEE.

Job Jobbed.

PS you'll need dichroic filters in front of the camera to reject the jamming from street lamps and led signs.

Good luck with this one. Will work fine in the lab... Not so well on the road. Exercise in why is left for the astute reader with  a Thorlabs DET-10A and a cheap digital scope in the passenger seat.

Then worry about hacking and jamming and older cars on the road.

There is a tiny segment of the 5 Ghz band allocated world wide for car to car.

Steve
« Last Edit: October 25, 2019, 05:47:02 am by LaserSteve »
"When in doubt, check the Byte order of the Communications Protocol, By Hand, On an Oscilloscope"

Quote from a co-inventor of the PLC, whom i had the honor of working with recently.
 

Offline TheUnnamedNewbie

  • Super Contributor
  • ***
  • Posts: 1211
  • Country: 00
  • mmwave RFIC/antenna designer
Ignoring the people who claim it is infeasible on practical reasons (While I agree it seems unlikely to be effective, part of research is looking at this regardless, because sometimes stuff pops up that turns out to be great after all):

Figure out if you have to do it in daylight or not. If so, you will have to look at SNR constraints, but it could still be doable - esp if you look at something 'spread spectrum' or rather high-redundancy.

An alternative approach to deal with low framrate issues: Is there any posibility you can look at driving the individual LEDs on a string of LEDs that make the tail light separately? You could try doing things like having low baudrate (so your camera can keep up) but high symbol depth information - if every LED can be tracked/modulated separately you could potentially encode a lot of information in one symbol. In a way, the properties of optical makes it so your lens acts like a extremely high gain antenna and you can look at it as a MIMO channel. Sure, you are limited to low symbol rates by the frame rate, but you will have much less issues with isolating one signal from another because you have such extreme directivity.

I think (but this is just me thinking out loud) that if you do this spatial modulation, people will notice it far less, since you still have an overall constant brightness (esp if you select the symbols in such a way that every symbol has a same average power).

Good luck with your bachelors thesis, sounds like an interesting project.


The best part about magic is when it stops being magic and becomes science instead

"There was no road, but the people walked on it, and the road came to be, and the people followed it, for the road took the path of least resistance"
 

Offline Someone

  • Super Contributor
  • ***
  • Posts: 5155
  • Country: au
    • send complaints here
I think (but this is just me thinking out loud) that if you do this spatial modulation, people will notice it far less, since you still have an overall constant brightness (esp if you select the symbols in such a way that every symbol has a same average power).
The constraint of a camera (unless its running at something beyond 200Hz, which just doesn't work in automotive space/cost limits) assures it will be noticeable to a human observer.
 

Offline damien22

  • Contributor
  • Posts: 30
  • Country: ch
I work in optics and this is definitely feasible (but not trivial).

Can it be a product on the market, perhaps, datalink would be wireless but headlights could send a simple ID to know who's the pair.

Definitely a cool project.

Modulation:
As stated, light are often controlled using PWM. However, things like the frequency can be modulated on a let's say 5% range step without influencing the PWM duty cycle and brightness of the LED. So the light intensity can still be regulated, and information sent.

Demodulation:
Here you will need some high response time sensors like photodiode and passive electronics to actually isolate the signal you are looking for. If the PWM frequency is on the 20kHZ range +/- modulation you would need to have a specific pass filter / amplifier for this frequency range and then some sort of MCU to define the exact frequency.

Camera definitely won't work.

Light pollution:
The receiver would need to have specific optics to filter the sunlight. A narrow bandpass filter in the 450nm~ or so that white LED usually exhibit will give a good signal VS sunlight. You would also probably want some lenses to focus the light coming from a specific direction.


Note that white LEDs have a response time due to the phosphor and that needs to be accounted for. Although if you have bandpass at 450nm that shouldn't be a concern.
 

Offline TheUnnamedNewbie

  • Super Contributor
  • ***
  • Posts: 1211
  • Country: 00
  • mmwave RFIC/antenna designer
I think (but this is just me thinking out loud) that if you do this spatial modulation, people will notice it far less, since you still have an overall constant brightness (esp if you select the symbols in such a way that every symbol has a same average power).
The constraint of a camera (unless its running at something beyond 200Hz, which just doesn't work in automotive space/cost limits) assures it will be noticeable to a human observer.

I am not so certain. I wonder if, when not directly in field of view, and only lightly modulated, say a few percent (and as I mentioned - in such away that the average power of the entire light is constant) people will really notice. It is not like you are hard switching on-and-off a single lightsource in a repetitive way.
The best part about magic is when it stops being magic and becomes science instead

"There was no road, but the people walked on it, and the road came to be, and the people followed it, for the road took the path of least resistance"
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf