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

Start-up to bring cheap, compact LIDAR to production cars by 2022!!

diplomat33

Average guy who loves autonomous vehicles
Aug 3, 2017
8,401
11,727
Terre Haute, IN USA
With Lidar, you still need to solve vision. Without Lidar you still need to solve vision. That is why Musk says its a fools errand.

True. Although to be more accurate, with lidar, you need less vision and without lidar, you need more vision.

I know lidar is not needed and Tesla can do fsd without lidar. But there is still a part of me that likes lidar and thinks it would still be nice for redundancy.
 

CarlK

Active Member
Mar 23, 2013
1,919
1,402
SF Bay Area
Keep in mind priority of start ups is always to convince investors not customers. For that nothing is too much for the unverifiable product spec and pricing info.

As for the Lidar debated I tend to believe it's not useful when the person who originally developed the Lidar based system says so.

 
  • Love
Reactions: StealthP3D

EVNow

Well-Known Member
Sep 5, 2009
10,694
30,873
Seattle, WA
I know lidar is not needed and Tesla can do fsd without lidar. But there is still a part of me that likes lidar and thinks it would still be nice for redundancy.
Or some other sensor that is not low fidelity, optical range. Get something else that can truly complement vision - getting information that is actually missing from vision.
 

CarlK

Active Member
Mar 23, 2013
1,919
1,402
SF Bay Area
Or some other sensor that is not low fidelity, optical range. Get something else that can truly complement vision - getting information that is actually missing from vision.

It could replace radar if it becomes as cheap (likely will never happen) although you don't really need to or want to replace radar which works better in foul weather.
 

Sean Wagner

Member
Mar 2, 2016
612
2,398
-
Or some other sensor that is not low fidelity, optical range. Get something else that can truly complement vision - getting information that is actually missing from vision.

That's what the brain does - supplying context by integrating data over time, thus helping to focus attention on anomalies.
 

diplomat33

Average guy who loves autonomous vehicles
Aug 3, 2017
8,401
11,727
Terre Haute, IN USA
Or some other sensor that is not low fidelity, optical range. Get something else that can truly complement vision - getting information that is actually missing from vision.

Like what? The only sensors that we have are cameras, radar, lidar or ultrasonics. You have to pick one of those. Also, vision offers the most information. There is no information that is missing from vision. That is why Tesla is so attracted to using cameras. The tricky part is that extracting the information from vision is more complex.
 

EVNow

Well-Known Member
Sep 5, 2009
10,694
30,873
Seattle, WA
Like what? The only sensors that we have are cameras, radar, lidar or ultrasonics. You have to pick one of those. Also, vision offers the most information. There is no information that is missing from vision. That is why Tesla is so attracted to using cameras. The tricky part is that extracting the information from vision is more complex.
Better Radar/Sonar, IR camera ?

By the time, FSD is good enough to start worrying about sensor deficiencies, they can pick the best available at that time. Currently it is a distraction.
 

diplomat33

Average guy who loves autonomous vehicles
Aug 3, 2017
8,401
11,727
Terre Haute, IN USA
Better Radar/Sonar, IR camera ?

By the time, FSD is good enough to start worrying about sensor deficiencies, they can pick the best available at that time. Currently it is a distraction.

I know Tesla can't add lidar now. They are committed to the current sensors. And I am sure that Tesla will get good city self-driving with cameras.

Honestly, I think I just get a little impatient with FSD. I see other companies who did start with lidar and who are ahead of Tesla in city self-driving and I wonder where Tesla would be if they had gone with lidar from the get-go. lidar probably would have allowed Tesla to start sooner on city self-driving, plus lidar probably would have solved the "dancing cars" problem, reduced phantom braking etc...
 

EVNow

Well-Known Member
Sep 5, 2009
10,694
30,873
Seattle, WA
I know Tesla can't add lidar now. They are committed to the current sensors. And I am sure that Tesla will get good city self-driving with cameras.

Honestly, I think I just get a little impatient with FSD. I see other companies who did start with lidar and who are ahead of Tesla in city self-driving and I wonder where Tesla would be if they had gone with lidar from the get-go. lidar probably would have allowed Tesla to start sooner on city self-driving, plus lidar probably would have solved the "dancing cars" problem, reduced phantom braking etc...

I'm fairly sure the reason Tesla didn't start with Lidar was that it didn't fir their strategy.

The strategy is
- Put the hardware on all cars
- Collect data and statistics to advance AP/FSD

Lidar simply didn't fit in this strategy. You can't put Lidar onobstrusively on commericial cars (esp. ones that are meant to be sleek and appeal to people who will pay 100k for them). And of course they cost a lot.

Now, if all of a sudden if we have inexpensive lidars that can be hidden like the cameras - it would be possible to put them in cars consistent with strategy - but Tesla will have a problem with 500k cars they have sold without lidars. So, at this point Tesla won't put lidars unless they think it is impossible to get to FSD without lidar (which is obviously not correct, given we all drive with just vision).

Basically the lidar ship has sailed, better to just concentrate on vision.
 
  • Like
Reactions: diplomat33

CarlK

Active Member
Mar 23, 2013
1,919
1,402
SF Bay Area
Better Radar/Sonar, IR camera ?

By the time, FSD is good enough to start worrying about sensor deficiencies, they can pick the best available at that time. Currently it is a distraction.

Definitely. Steph Curry recently did a lasik surgery to correct a vision problem he had the whole life. That's after he spent two decades to train his brain to play good basketball. I could have ten lasik surgeries and implant two extra eyes and still would not be able to play high school level basketball. Those who are so hung up on sensors seem to have a hard time to see (sorry about the pun) how this whole thing works.
 

acoste

Member
Nov 1, 2018
742
715
California
I'm fairly sure the reason Tesla didn't start with Lidar was that it didn't fir their strategy.

The strategy is
- Put the hardware on all cars
- Collect data and statistics to advance AP/FSD

Lidar simply didn't fit in this strategy. You can't put Lidar onobstrusively on commericial cars (esp. ones that are meant to be sleek and appeal to people who will pay 100k for them). And of course they cost a lot.

Now, if all of a sudden if we have inexpensive lidars that can be hidden like the cameras - it would be possible to put them in cars consistent with strategy - but Tesla will have a problem with 500k cars they have sold without lidars. So, at this point Tesla won't put lidars unless they think it is impossible to get to FSD without lidar (which is obviously not correct, given we all drive with just vision).

Basically the lidar ship has sailed, better to just concentrate on vision.


Vision is not perfect and to use it as a reason not to have additional sensors is incorrect. Why not have something better and safer as well? Radar is not vision (although it is EM waves as well), yet Elon puts them in the car. And it is the cost why he left lidar out, not bc they are useless.

Also a human driver has no redundancy. Once we have technology capable of self driving, we can add redundancy. That's another improvement.

MobilEye has reached good results with cameras only, as distance measuring devices. But they also add that only "in good weather". And they are planning to add redundancy where needed using radars and lidars.
It is possible to have improved radar performance. But lidar's resolution is a lot higher and can read human body or a bicyclist's movement which helps with predictions.

As for your 90mph comment about lidar:
It takes about 300 feet to stop from 90 mph. Many lidars have the range of 120m. However range is defined at a specific reflectivity, like 10%. Cars and road barriers usually have larger than 10% reflectivity so they can be spotted earlier. This guy here can detect cars up to 280m.

LiDAR System Raises Safety-Performance Bars
 
  • Disagree
Reactions: plumazul

EVNow

Well-Known Member
Sep 5, 2009
10,694
30,873
Seattle, WA
Why not have something better and safer as well?
Because it won't let you collect large amount of data or help statistically prove how better FSD is compared to humans. In other words, for Tesla's strategy, Lidars are not practical.

Look at it this way.

- You either have lidar but very limited data and have to rely on simulation
- No lidar but a lot of data and driven miles

You can't have both. So, instead of "Why not have something better and safer as well?" - how about "Why not have much more data and real data too?".
 
  • Like
Reactions: CarlK

CarlK

Active Member
Mar 23, 2013
1,919
1,402
SF Bay Area
Because it won't let you collect large amount of data or help statistically prove how better FSD is compared to humans. In other words, for Tesla's strategy, Lidars are not practical.

Look at it this way.

- You either have lidar but very limited data and have to rely on simulation
- No lidar but a lot of data and driven miles

You can't have both. So, instead of "Why not have something better and safer as well?" - how about "Why not have much more data and real data too?".

Huge deep learning data is one thing but Lidar also does not have the resolution that could allow the NN to do things like to observe small movements of cars and pedestrians to predict their intentions. If you listen to that Levandowski interview this is the problem needs to be resolved at the moment to unlock the self driving technology. I think it makes a lot of sense. It's the key to everything people have concern with from 4 way stop to pedestrians on the curb to cars in the next lane trying to move.
 

EVNow

Well-Known Member
Sep 5, 2009
10,694
30,873
Seattle, WA
Huge deep learning data is one thing but Lidar also does not have the resolution that could allow the NN to do things like to observe small movements of cars and pedestrians to predict their intentions. If you listen to that Levandowski interview this is the problem needs to be resolved at the moment to unlock the self driving technology. I think it makes a lot of sense. It's the key to everything people have concern with from 4 way stop to pedestrians on the curb to cars in the next lane trying to move.
Yes, as Levandowski says Waymo never had a disengagement because a sensor failed. It was because they had problem with figuring out what the various actors on the scene would do next.
 

acoste

Member
Nov 1, 2018
742
715
California
Because it won't let you collect large amount of data or help statistically prove how better FSD is compared to humans. In other words, for Tesla's strategy, Lidars are not practical.

Look at it this way.

- You either have lidar but very limited data and have to rely on simulation
- No lidar but a lot of data and driven miles

You can't have both. So, instead of "Why not have something better and safer as well?" - how about "Why not have much more data and real data too?".

I get that due to economics there is no chance of having both lidar and large amount of data at the same time. And therefore it is perfectly fine to use camera only for now for data collection.
And one can decide later down the road if more sensors are needed for autonomous driving. Saying however that cameras will be enough for autonomous driving is just a wild prediction until someone gets there.
Elon says lidar is fools errand due to business reasons. How could he sell "FSD" now otherwise? It's not a technical decision.
 
  • Like
Reactions: caligula666

acoste

Member
Nov 1, 2018
742
715
California
Huge deep learning data is one thing but Lidar also does not have the resolution that could allow the NN to do things like to observe small movements of cars and pedestrians to predict their intentions. If you listen to that Levandowski interview this is the problem needs to be resolved at the moment to unlock the self driving technology. I think it makes a lot of sense. It's the key to everything people have concern with from 4 way stop to pedestrians on the curb to cars in the next lane trying to move.

Why can't observe small movements??? I can clearly see arms and legs.

 

EVNow

Well-Known Member
Sep 5, 2009
10,694
30,873
Seattle, WA
Saying however that cameras will be enough for autonomous driving is just a wild prediction until someone gets there.
Elon says lidar is fools errand due to business reasons. How could he sell "FSD" now otherwise? It's not a technical decision.
You can actually make the same argument about lidar.

Lidar is being used only to get to some kind of geofenced L4 quickly to get VC money.

Vision only FSD is not wild speculation. It is just "first principles". If humans can do it with vision only, theoretically we should be able to do it with NN as well.
 
  • Like
Reactions: CarlK and plumazul

Products we're discussing on TMC...

About Us

Formed in 2006, Tesla Motors Club (TMC) was the first independent online Tesla community. Today it remains the largest and most dynamic community of Tesla enthusiasts. Learn more.

Do you value your experience at TMC? Consider becoming a Supporting Member of Tesla Motors Club. As a thank you for your contribution, you'll get nearly no ads in the Community and Groups sections. Additional perks are available depending on the level of contribution. Please visit the Account Upgrades page for more details.


SUPPORT TMC
Top