Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Sense Photonics brings its new LIDAR to market

This site may earn commission on affiliate links.

diplomat33

Average guy who loves autonomous vehicles
Aug 3, 2017
12,555
18,383
USA
Sense Photonics is bringing its new solid state flash LIDAR to market. They are the more compact, non spinning type. Sense is focusing on wide angle and short range. They cost $2900 a piece.
Sense Solid-State Flash LiDAR - Sense Photonics
Sense Photonics brings its fancy new flash lidar to market – TechCrunch

I am a big LIDAR fan. I know LIDAR has some strengths that make them a good sensor for autonomous driving. I am very happy to see the new LIDAR is moving away from the bulky spinning types and becoming more compact. But Sense's LIDAR still seems a bit expensive. You would need at least 4 units for an autonomous car which means a total cost of over $10k per car. That would probably ruin your profit margin. Not sure if buying in bulk would bring the cost down enough.

I could see it maybe working for ride hailing company that intends to build a big enough fleet of robotaxis and then hopes to recoup the costs over time from the service, especially with the savings from not needing to pay human drivers and cheaper insurance. But I can see why Tesla is not interested in LIDAR. There is no way Tesla could afford to put LIDAR on hundreds of thousands of cars. Maybe Tesla would consider LIDAR on just the robotaxis if the price comes down more?

Thoughts?
 
  • Informative
Reactions: jebinc
...Maybe Tesla would consider LIDAR on just the robotaxis if the price comes down more?...

I haven't heard the cost is among the reasons for Tesla to refuse LIDAR.

Tesla's reasons have been that LIDAR is not appropriate for Self-Driving. Not only that it does not help but also it's a crutch.

Thus, I don't see how Tesla would be willing to take something cheap or free just to slow down its progress.
 
  • Funny
Reactions: Daniel in SD
Not only that it does not help but also it's a crutch.
Crutches do help. If they didn't then why would people use them? o_O

I wonder if the time of flight camera technology that's starting to show up in smartphones will make its way to cars. It's basically the same idea as LIDAR but instead of a scanning laser or array it just sends out a flash of infrared light and uses a time of flight camera to get depth information. Seems like it could be very useful for autopark and maneuvering in tight spaces.
 
  • Helpful
Reactions: diplomat33
I haven't heard the cost is among the reasons for Tesla to refuse LIDAR.

Tesla's reasons have been that LIDAR is not appropriate for Self-Driving. Not only that it does not help but also it's a crutch.

Thus, I don't see how Tesla would be willing to take something cheap or free just to slow down its progress.

I believe cost is the REAL reason and certainly a very legitimate reason. The official reasons that Elon has given are basically BS. LIDAR has good range, provides high resolution mapping at the speed of light and works great in both the dark and in bright light where cameras might be blinded. So LIDAR is definitely very useful and appropriate for self-driving. There must be a good reason why every company working on autonomous driving is using LIDAR and showing better results than Tesla so far.

So yeah cost could definitely be a factor. But saying LIDAR is not useful to autonomous driving. Sorry, not buying it.

I wonder if the time of flight camera technology that's starting to show up in smartphones will make its way to cars. It's basically the same idea as LIDAR but instead of a scanning laser or array it just sends out a flash of infrared light and uses a time of flight camera to get depth information. Seems like it could be very useful for autopark and maneuvering in tight spaces.

I think you just described flash LIDAR, what Sense Photonics is doing.
 
...So yeah cost could definitely be a factor. But saying LIDAR is not useful to autonomous driving. Sorry, not buying it...

I don't think Tesla is anti-LIDAR. Elon said he does use LIDAR for SpaceX but he just doesn't think it's an appropriate use in Self-driving cars.

The reason everyone else is using LIDAR is that they haven't been able to figure out how to get the cameras to work for them.
 
Yeah, it does seem to be exactly the same thing. For some reason I thought flash LIDAR used some sort of optics to project a grid of laser beams but it appears that it's just a single wide cone. Not sure why it costs so much more than what they're using on smartphones. I guess it has longer range.

Yeah the range is higher. The article says the range is up to 40 m which I am guessing is more than what a smart phone can do.
 
I don't think Tesla is anti-LIDAR. Elon said he does use LIDAR for SpaceX but he just doesn't think it's an appropriate use in Self-driving cars.

The reason everyone else is using LIDAR is that they haven't been able to figure out how to get the cameras to work for them.

The fact that Elon is not anti-LIDAR at SpaceX does in no way exclude the possibility that he is anti-LIDAR at Tesla for cost reasons.

I mean any (realistic) cost of LIDAR is irrelevant in space where everything costs millions anyway. This is not the case in consumer cars.

That said, I wouldn’t be surprised if Elon isn’t anti-LIDAR on principle by now at Tesla. He has doubled down on it for so long, there must be some psychological ”skin in the game” by now... Time will tell if it is another Alien Dreadnought mistake or something he can fake until they make it.
 
The reason everyone else is using LIDAR is that they haven't been able to figure out how to get the cameras to work for them.

I am not sure that is entirely correct. Waymo and others use cameras as part of their autonomous driving. Mobileye has even built quite a lot of camera neural nets. I think the main reason they all use LIDAR is for the extra redundancy and reliability. Their autonomous driving philosophy is to maximize sensor redundancy in order to get the highest reliability possible. Getting data from cameras, radar and LIDAR reduces the chances of the car missing something than if you just rely on cameras and radar alone. LIDAR, combined with HD maps, also allows you the determine the position of the car in its environment with very high accuracy which helps the car navigate better. You can duplicate this with just cameras but its harder. I think the companies probably feel that since LIDAR and HD maps work so well in localizing the car in its environment with very high accuracy, why not use that method? And again, for redundancy sake, even if your cameras do the job, why not have LIDAR and HD maps double check the cameras and make sure you have the best localization possible?

In fact, I suspect that the deaths and accidents of Teslas might have been avoided with LIDAR because LIDAR might have detected the semi or the crash guards when the cameras obviously missed them. I also suspect that LIDAR could probably help solve some of the problems like phantom braking that Tesla is having. Maybe Tesla will solve those problems without LIDAR but LIDAR could perhaps help solve them sooner and more reliably.

I understand Elon's approach is to develop camera vision that is so good that the car will not need LIDAR or HD maps in order to navigate. That is an aspirational goal. Certainly, if and when Teslas achieves that goal, they would have an autonomous driving tech that is simpler, cheaper and more elegant than what the other companies have. My point is simply that LIDAR would still be a good idea because it would provide extra reliability while Tesla works to get the better camera vision and even beyond.
 
Last edited:
...In fact, I suspect that the deaths and accidents of Teslas might have been avoided with LIDAR because LIDAR might have detected the semi or the crash guards when the cameras obviously missed them...

I don't think Tesla cameras are used for those fatal accidents just yet. It's more like radar that is currently responsible to figure out which are harmless overpass to go under and which are stationary obstacles right in front of the car that it should not allow the car to crash into.

Tesla's camera system is there but it has not achieved the competency to tell the brake system to automatically brake for those stationary obstacles just yet.

It's just like the camera system is there but it has not achieved the competency to tell the brake system to automatically brake for those red lights just yet.

So Tesla's camera's advantage is still a theory and has not been proven just yet.

...My point is simply that LIDAR would still be a good idea because it would provide extra reliability while Tesla works to get the better camera vision and even beyond.

Not everyone would subscribe to the above idea: Instead of using LIDAR as an incremental bridge, Tesla skips that step completely.

So who is right?

We don't know yet because none of this is proven.

It looks like Waymo is leading but if I as an average consumer can buy one with good money, it doesn't have the general practical use for me because it would get lost in a construction zone in Los Angeles freeway that the cars are redirected to the physically wrong side of the freeway. And that the redirection keeps changing as it's now in Bakersfield as well!

Tesla Autopilot can handle construction redirections very well but it still has the problem of fatal accidents in good road conditions with no constructions.

But if I have to choose, I would choose the flexible system from Tesla without LIDAR rather than the inflexible ones from Waymo with its sophisticated LIDAR.
 
Not everyone would subscribe to the above idea: Instead of using LIDAR as an incremental bridge, Tesla skips that step completely.

...

But if I have to choose, I would choose the flexible system from Tesla without LIDAR rather than the inflexible ones from Waymo with its sophisticated LIDAR.

There is no reason to think, though, that Lidar would have to be:

a) an incremental bridge
b) cause an inflexible system

Just because Lidar may be used as part of either in some project or another, does not mean it couldn’t be a useful part of a very flexible, non-transitional system.

It is a sensor that provides certain unique benefits in object and distance detection — even compared to vision. Even if we assume perfect camera vision, Lidas does see beyond visual range and its characteristics differ from cameras sufficiently to generate real redundancy in various hindered visibility scenarios. The ranging is also very accurate.

I agree it is an open question which sensors eventually can and will be used and to what effect in autonomous driving. It will probably be a mix of various sensor suites and technologies, and probably not the same in all systems...
 
Last edited:
I thought it was fairly well done. After all, the NOVA series has been on the air for something like 46 years...

Yes, it was very well done. It certainly did a great job of laying out both the potential rewards and risks of autonomous driving. It also did a great job of explaining how autonomous tech works and what the current state of autonomous driving is.

My critique of the video is that it seemed a bit too pessimistic in places, specifically on when L5 will happen. It says that L5 will probably take a very long time (like 30-40 years) because experts are doubtful that autonomous cars can do better than the human average of 1 crash death per 100 million miles. But I think the tech will probably get there faster. The current autonomous tech might not be able to do better than 1 crash death per 100 million miles but technology progresses quickly and it is hard to tell what the technology will be just 5 years from now. 5-10 years from now, the technology might very be good enough for L5 autonomy.

Also, the video seems to ignore that we can get benefits from L4 autonomy. And we already have L4 prototypes now on the road in some limited areas. So we can get some usefulness from autonomous driving much sooner. We don't need to wait for L5 to see benefits.
 
  • Like
Reactions: jebinc