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

[diplomat33]

"Lidar startup Luminar announced today that it’s received $100 million in new funding, as the company also introduced Iris, an autonomous platform that it expects to bring to production vehicles by 2022 — for less than $1,000.

Luminar says its new Iris platform will launch commercially on production vehicles starting in 2022. It boasts that Iris is “the first sensing platform to exceed the essential performance, safety, cost, and auto-grade requirements needed to deliver Level 3 and 4 autonomy to consumers."
Lidar startup Luminar gets $100M in funding, sub-$1,000 Iris platform set for production cars by 2022 - Electrek

I like the price and I also really like that the LIDAR is not bulky either. it seems like a great solution. If it works, there really won't be any excuse not to add LIDAR on FSD cars.

This could really help bring full self-driving to mass market production cars. I could see companies like Waymo and Cruise using Luminar to bring their FSD to cars in a more compact package.

I am also intrigued by the idea in the article of the LIDAR set as an after market addition. Would it be possible to add it to a Tesla as an after market solution to improve reliability if Tesla does not?

 

[Tam]

"...the first sensing platform to exceed the essential performance, safety, cost, and auto-grade requirements needed to deliver Level 3 and 4 autonomy to consumers...

It's meaningless without the specs.

I think Velodyne HDL-64E is still the gold standard that has a range of 120 meters or just over 390 feet which is slightly longer than a football field. It's about $100,000 though.

 

[Puddles]

Assuming it’s not vaporware it will put Elon’s claim that Lidar is “a fools errand” to the test. I’d lay even money on the two outcomes at this point.

 

[plumazul]

I like the price and I also really like that the LIDAR is not bulky either. it seems like a great solution. If it works, there really won't be any excuse not to add LIDAR on FSD cars.

What exactly is LIDAR a solution to?

 

[diplomat33]

What exactly is LIDAR a solution to?

LIDAR is a high precision sensor for detecting objects around the car. LIDAR is useful for "painting" a highly precise map of the environment around the car. It can also add redundancy in case your cameras or other sensors fail. For example, if Teslas had front LIDAR years ago, Brown and others probably would not have died in those traffic accidents since LIDAR would have detected the stopped car or the semi pulling out in front when the cameras and radar did not.

A big reason why Tesla has not used LIDAR is that they are too expensive and too bulky. But if Luminar is successful at making a LIDAR set for just a $1000 that is also compact, then it would negate those reasons not to use it.

 

[Puddles]

My understanding of LIDAR is that it requires moving parts. I think it can only become realistic for mass usage if it is made to be solid state. I'm sure this is the focus of attention right now! I don't see why the movement couldn't come in the form of vibrations, magnified by a mirror, to simplify the mechanism. Or maybe I've been to too many laser light shows.

 

[Tam]

...if Teslas had front LIDAR years ago, Brown and others probably would not have died in those traffic accidents since LIDAR would have detected the stopped car or the semi pulling out in front when the cameras and radar did not...

That's in theory but not proven in reality.

Have you ever wondered why there's no LIDAR test video released on Youtube for its cars running at 90MPH and would successfully brake for a stationary obstacle?

Why do you think UBER had to turn off its automation system when it clearly had LIDAR in its system?

Report: Uber self-driving team was preparing for CEO demo before fatal crash

 

[diplomat33]

That's in theory but not proven in reality.

Have you ever wondered why there's no LIDAR test video released on Youtube for its cars running at 90MPH and would successfully brake for a stationary obstacle?

Can you give one reason why LIDAR would not work at 90 mph? I can't think of any.

Why do you think UBER had to turn off its automation system when it clearly had LIDAR in its system?

Report: Uber self-driving team was preparing for CEO demo before fatal crash

The article explains that UBER made the decision to turn off the automatic emergency braking because of really bad phantom braking and they wanted the demo to be smooth. But the bad phantom braking was due to UBER's bad software. It had nothing to do with LIDAR.

 

[Tam]

...The article explains that UBER made the decision to turn off the automatic emergency braking because of really bad phantom braking and they wanted the demo to be smooth. But the bad phantom braking was due to UBER's bad software. It had nothing to do with LIDAR.

That's the problem of theory and implement the theory to real life.

Theory works without a boss. Now, in real life, people have a boss that they have to serve.

It has nothing to do with LIDAR but everything with reality!

 

[diplomat33]

That's the problem of theory and implement the theory to real life.

Theory works without a boss. Now, in real life, people have a boss that they have to serve.

It has nothing to do with LIDAR but everything with reality!

Sure, a car can still hit something even if it has LIDAR, if you willfully turn off AEB. But that is because you are deliberately turning off the feature designed to stop the car. Of course, it will hit something! That has no bearing on whether LIDAR is capable of helping stop the car or not.

 

[Tam]

...Can you give one reason why LIDAR would not work at 90 mph? I can't think of any...

LIDAR can work very well in slow speed. However, when you are talking about high speed, it might be out of range for LIDAR.

Take the best range that we have now: Velodyne HDL-64E that has a range of 120 meters or just over 390 feet which is slightly longer than a football field.

90 MPH x 5,280 feet = 475,000 feet per hour or 132 feet per second.

In 3 seconds, 90 MPH would travel 396 feet or a length of football field.

So by the time LIDAR can detect an obstacle a football field away, it has to decide instantly whether to do a complete halt or delaying a fraction later which might be too late because a complete halt may take several hundreds of feet from the hard braking.

 

[Tam]

Sure, a car can still hit something even if it has LIDAR, if you willfully turn off AEB. But that is because you are deliberately turning off the feature designed to stop the car. Of course, it will hit something! That has no bearing on whether LIDAR is capable of helping stop the car or not.

It's the same when other cars company say they can beat Tesla with further range and much faster-charging rate.

If Tesla can do it, others can do it better!

Looks good in theory, but to implement the theory to reality is a completely different story.

 

[diplomat33]

LIDAR can work very well in slow speed. However, when you are talking about high speed, it might be out of range for LIDAR.

Take the best range that we have now: Velodyne HDL-64E that has a range of 120 meters or just over 390 feet which is slightly longer than a football field.

90 MPH x 5,280 feet = 475,000 feet per hour or 132 feet per second.

In 3 seconds, 90 MPH would travel 396 feet or a length of football field.

So by the time LIDAR can detect an obstacle a football field away, it has to decide instantly whether to do a complete halt or delaying a fraction later which might be too late because a complete halt may take several hundreds of feet from the hard braking.

That's why you have sensor fusion. The front of the car usually has cameras, radar and lidar.

camera: 250 m
radar: 160 m
lidar: 120 m

lidar is your last line of defense. The camera with the longest range will hopefully detect the stopped car first. If not, then the radar will pick it up next. Lastly, the lidar will pick up the stopped car. That's 3 sensors to detect the stopped car.

 

[Tam]

...That's why you have sensor fusion...

Completely agreed. I have no problem with that in theory!

 

[Uncle Paul]

They are going to borrow $100,000,000.00 and then sell the LIDAR for cheap. Not happening. They will sell for what ever the market will bear. Maybe down the road they will hit their projected pricing, but not for many years.

 

[plumazul]

LIDAR is a high precision sensor for detecting objects around the car. LIDAR is useful for "painting" a highly precise map of the environment around the car.

?? You mean like the Sun or nighttime lighting?

A big reason why Tesla has not used LIDAR is that they are too expensive and too bulky. But if Luminar is successful at making a LIDAR set for just a $1000 that is also compact, then it would negate those reasons not to use it.

I think the BIGGER reason Tesla has not used LIDAR is because Elon thinks it's worthless(adds nothing).

 

[Uncle Paul]

On the other hand, it LIDAR gets cheap and effective, I imagine that Tesla will integrate it into their sensor suite...no biggie.

 

[diplomat33]

I think the BIGGER reason Tesla has not used LIDAR is because Elon thinks it's worthless(adds nothing).

Well, the reason Elon thinks it adds nothing is because Elon is banking on camera vision being better than lidar. Basically, if your vision neural net is good enough that your car sees and understands everything around it then you don't need lidar. It's the "human principle". Humans have eyes and a brain that are good enough that we can drive with just what we see. Elon hopes to do the same with cameras. The tricky part for Tesla is that for that approach to work, the vision neural net has to be amazing. It not just has to read traffic signs and such but it has to be able to understand everything it sees and even figure out distance just from pixels. And it takes a lot of machine learning to get there. But in theory, once Tesla has a complete vision neural net, then they will be able to do full self-driving without lidar. Other companies like Waymo have been able to progress faster towards full self-driving by using lidar. They only need partial vision and can use lidar for the rest.

 

[EVNow]

Assuming it’s not vaporware it will put Elon’s claim that Lidar is “a fools errand” to the test. I’d lay even money on the two outcomes at this point.

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.

 

[EVNow]

even figure out distance just from pixels.

That is solved - both at Tesla and at Google. I bet Waymo has a project now that is working on FSD without Lidar.