Fixed object detection in Autopilot/FSD (out of main)

[KarenRei]

Maybe not topical, but IMO important and definitely related to Tesla's competitors. I have a strong belief that LIDAR should be banned due to the risk it poses to vision -- both of humans and animals.

Roughly speaking, the argument in favor of LIDAR safety goes that the lasers are kept to low enough levels to avoid problems. However, I have not found anything convincing that this has been thoroughly studied or even takes into account individual variance. For example, what is the expected rate for eye damage with currently approved LIDAR systems? Does it assume an ideal state with only one LIDAR unit in use? How rigorous are the protocols for determining whether or not eye damage has occurred?

My searching has not been particularly fruitful so maybe I have missed something and LIDAR has been shown to be safe. One article[1] I did find indicates that studies are "on-going" and that there is substantial risk to overlooking eye damage due to compensation concealing blind spots until the damage is severe. I am concerned that, unless checked, use of LIDAR on public roads will not only have an undesirable impact on animals, but also result in a wave of severe eye damage.

Am I wrong? I'd love to be, and I admit I don't have much data to work with in drawing my conclusions. But as it stands, I think automotive LIDAR should be banned.

1) Lasers and Eye Safety: Are Lasers Dangerous? | SemiNex

I've long had the exact same concern - that the parameters for prediction of eye damage risk are based on far too naive of assumptions.

 

[Chocochip]

Maybe not topical, but IMO important and definitely related to Tesla's competitors. I have a strong belief that LIDAR should be banned due to the risk it poses to vision -- both of humans and animals.

Roughly speaking, the argument in favor of LIDAR safety goes that the lasers are kept to low enough levels to avoid problems. However, I have not found anything convincing that this has been thoroughly studied or even takes into account individual variance. For example, what is the expected rate for eye damage with currently approved LIDAR systems? Does it assume an ideal state with only one LIDAR unit in use? How rigorous are the protocols for determining whether or not eye damage has occurred?

My searching has not been particularly fruitful so maybe I have missed something and LIDAR has been shown to be safe. One article[1] I did find indicates that studies are "on-going" and that there is substantial risk to overlooking eye damage due to compensation concealing blind spots until the damage is severe. I am concerned that, unless checked, use of LIDAR on public roads will not only have an undesirable impact on animals, but also result in a wave of severe eye damage.

Am I wrong? I'd love to be, and I admit I don't have much data to work with in drawing my conclusions. But as it stands, I think automotive LIDAR should be banned.

1) Lasers and Eye Safety: Are Lasers Dangerous? | SemiNex

Not sure about eye safety, but I remember seeing a report of camera sensors getting permanently damaged by LIDAR, a photojournalist shooting some LIDAR-equipped cars noticed this. Possibly the result of the camera lens focusing the beam, although I’m not quite sure (a laser beam, by definition, is already a quite fine and intense light beam). That effect is arguably suboptimal when some other cars on the road use either vision-based autonomous features or dash cams, or both. It may be the case that LIDAR systems can be adjusted such that the emitted laser beam intensity is adjusted to below damage-causing thresholds, but… it feels like adding a problem to fix when there’s no need for one to begin with.

 

[jerry33]

Not sure about eye safety, but I remember seeing a report of camera sensors getting permanently damaged by LIDAR, a photojournalist shooting some LIDAR-equipped cars noticed this. Possibly the result of the camera lens focusing the beam, although I’m not quite sure (a laser beam, by definition, is already a quite fine and intense light beam). That effect is arguably suboptimal when some other cars on the road use either vision-based autonomous features or dash cams, or both. It may be the case that LIDAR systems can be adjusted such that the emitted laser beam intensity is adjusted to below damage-causing thresholds, but… it feels like adding a problem to fix when there’s no need for one to begin with.

The problem is that to be effective LIDAR needs to be as powerful as possible, so it's presumably tuned to just below the level thought to the eye damaging. Besides the sensitivity variation of people, and possible incorrect adjustment, there's also the multiple beam from different cars issue.

 

[PhaseWhite]

The problem is that to be effective LIDAR needs to be as powerful as possible, so it's presumably tuned to just below the level thought to the eye damaging. Besides the sensitivity variation of people, and possible incorrect adjustment, there's also the multiple beam from different cars issue.

Even if LIDAR only damaged camera sensors that sounds like a major safety concern to all the other cars on the road that will rely on camera sensors.


Full Page Reload

 

[humbaba]

Not sure about eye safety, but I remember seeing a report of camera sensors getting permanently damaged by LIDAR, a photojournalist shooting some LIDAR-equipped cars noticed this. Possibly the result of the camera lens focusing the beam, although I’m not quite sure (a laser beam, by definition, is already a quite fine and intense light beam). That effect is arguably suboptimal when some other cars on the road use either vision-based autonomous features or dash cams, or both. It may be the case that LIDAR systems can be adjusted such that the emitted laser beam intensity is adjusted to below damage-causing thresholds, but… it feels like adding a problem to fix when there’s no need for one to begin with.

The camera issue you describe was most likely this one: Man says CES lidar’s laser was so powerful it wrecked his $1,998 camera

Here is a spec sheet for a LIDAR unit that claims "[it] is safe to look at with the unaided eye", but then says doing so is a bad idea. And that seems to be a toy unit with a maximum range of 40 meters. Just as a first take, ~40m isn't enough for a safe follow distance, much less "seeing" the situation ahead in order to take action in a timely fashion.

https://cdn.sparkfun.com/datasheets/Sensors/Proximity/lidarlite2DS.pdf

 

[Unpilot]

Even if LIDAR only damaged camera sensors that sounds like a major safety concern to all the other cars on the road that will rely on camera sensors.


Full Page Reload

Plus all the security camera's that are literally everywhere.

How Many Times Are You Caught on Security Camera per Day – Reolink Blog

 

[nativewolf]

Plus all the security camera's that are literally everywhere.

How Many Times Are You Caught on Security Camera per Day – Reolink Blog

Sounds like a good reason for Lidar. In fact, maybe I'll put lidar on my old beater

 

[StealthP3D]

Even if LIDAR only damaged camera sensors that sounds like a major safety concern to all the other cars on the road that will rely on camera sensors.


Full Page Reload

Now Tesla's competitors illogical insistence on using LIDAR is starting to make sense...

If you can't beat Tesla to the prize, make sure they never get there by filling the roads with camera destroying LIDAR!

 

[pz1975]

The camera issue you describe was most likely this one: Man says CES lidar’s laser was so powerful it wrecked his $1,998 camera

Here is a spec sheet for a LIDAR unit that claims "[it] is safe to look at with the unaided eye", but then says doing so is a bad idea. And that seems to be a toy unit with a maximum range of 40 meters. Just as a first take, ~40m isn't enough for a safe follow distance, much less "seeing" the situation ahead in order to take action in a timely fashion.

https://cdn.sparkfun.com/datasheets/Sensors/Proximity/lidarlite2DS.pdf

Good summary article on this topic. There is no consensus yet on eye safety and hopefully some proper studies will be done before the technology gets out there in the real world in large numbers (if it ever does).

https://www.laserfocusworld.com/blogs/article/14040682/safety-questions-raised-about-1550-nm-lidar

 

[humbaba]

Good summary article on this topic. There is no consensus yet on eye safety and hopefully some proper studies will be done before the technology gets out there in the real world in large numbers (if it ever does).

https://www.laserfocusworld.com/blogs/article/14040682/safety-questions-raised-about-1550-nm-lidar

Thanks for that, you are right, that does seem like a good summary. It doesn't make me feel any better about the situation and I wish senator Malarky and gang would take up this issue instead of bashing autopilot which has already been investigated once by the NHTSA and found to decrease accident rates by 40%.

 

[pz1975]

Thanks for that, you are right, that does seem like a good summary. It doesn't make me feel any better about the situation and I wish senator Malarky and gang would take up this issue instead of bashing autopilot which has already been investigated once by the NHTSA and found to decrease accident rates by 40%.

Interestingly the American Academy of Ophthalmology (AAO) has no mention of LIDAR in relation to ocular safety anywhere on their website. I wrote them a letter asking regarding this (see below) and will report back when I get a response. I agree this is a very important topic as we move forward.

>I am an ophthalmologist in British Columbia, Canada and am asking if the AAO has a current opinion or any ongoing studies/investigations happening regarding the safety of LIDAR on ocular health, especially as it pertains to its application in autonomous vehicles. There is a lot of research worldwide in developing autonomous vehicles and one of the major technologies being investigated is LIDAR to map objects around the cars to allow for safe autonomous driving.

From my reading in this topic, it appears that 905 nm LIDAR has been determined to be unsafe due to possible retina damage, while 1550 nm has so far been determined to be safe for the eye despite the higher power at that wavelength needed to obtain useful images for safe autonomous driving, and the possible damage to the cornea and lens due to absorption of this wavelength by those structures (the retina is safe with 1550 nm since the light at that wavelength is blocked by cornea, lens and vitreous).

I am concerned about this topic because autonomous driving will likely become commonplace in society one day (maybe in the next 3-10 years) and it is possible that LIDAR systems may become the dominant technology used. Other companies are testing other systems like regular cameras and radar but there has yet be a “winner” in the technology race to achieve true autonomous driving.

If LIDAR were to win this race in the future, it would be difficult at that point to pushback against its use for any possible safety reasons since hundreds of billions of dollars of potential profit will be at stake since it would redefine the entire transportation industry.

As an ophthalmologist, I think we need investigate this technology closely now while it is still in the research stage in order to proverbially “keep the genie in the bottle” if it does turn out to be a threat to ocular health. It would be too late if thousands and eventually millions of cars on the road have this technology and only then we start to see possible ocular damage such corneal burns/opacities and/or cataract (or other unknown pathologies) among our patients.

Thank you for your consideration.