Lucid unveils advanced driver assist features to compete with Autopilot

[MP3Mike]

It's less than $1000 per car on a car that costs around $100k. I doubt it will cut too much into the profit margins.

So you are saying that the Model 3, which starts at $35k, wouldn't get it?

Do you realize that the average sales price for a Tesla right now is ~$52k? So no, a Tesla doesn't cost around $100k. (Only their top end S&X performance models cost that much.)

 

[diplomat33]

So you are saying that the Model 3, which starts at $35k, wouldn't get it?

Do you realize that the average sales price for a Tesla right now is ~$52k? So no, a Tesla doesn't cost around $100k. (Only their top end S&X performance models cost that much.)

You are changing the topic. We were talking about Lucid, not Tesla. I was responding to your claim that Lucid could not afford it.

But the lidar costs between $500-$1000. I think the average Tesla (~$50k) could get the lidar too. I don't see a problem with a cheaper Tesla also getting it. I for one, would have happily paid an extra $1000 for a front lidar on my Model 3 LR.

 

[WarpedOne]

>> But the lidar costs between $500-$1000.

What a pile of BS... 1000$ may be the cost of that box with lidar inside, that does not a L3 vehicle make.

What is the cost of development dead-ends that lidar will make you go into?
It is a hammer and it will make you solve the problems as if they all are all nails.
Until you run into into a problem where mountains of burned cash still hasn't brought you to the other side.

This is that audi L3 BS talk all over again. It should be in customer hands in 2018.
Lucid is still just a hot air baloon, not even a lucid dream ...

Job #1: deliver a single customer car.

 

[Bladerskb]

Thanks for sharing. I think these sensors are awesome! And it will be powered by Mobileye so the computer power should be really good too. It looks to me like Lucid should be able to provide excellent highway L3.

Interesting. It looks like they did trim down the sensors a bit since an earlier diagram showed side and rear lidar coverage as well. But I like the front lidar. I think the front lidar should provide excellent reliability for those situations like a stopped vehicle or a white semi crossing in front.

I find it very interesting that even with all these sensors, Lucid doesn't consider the current sensor/hardware good enough for L4 and therefore only plans to do L3 right now. I am guessing it is because it lacks side and rear lidar. So for city driving where you might have a lot of objects all around you, the sensors lack a bit of redundancy on the sides and rear. You don't want a situation where the car does not see cross traffic at an intersection because the side cameras misses it. Hence, why they don't want to attempt L4 on this sensor suite.

I think the biggest problem with offering L4 in this current sensor suite they have is compute redundancy and sensor cleaning. Because their current sensor suite is sufficient for L4 highway since its multi modal (360 cameras & 360 radars) and triple modal redundancy in the front.

 

[Bladerskb]

Well, as people said... it looks good on paper.

I have not too much doubt that in perfect condition (freeway with no obstacle) everything would work as describe...

but in real life it would have the same issues that the Tesla FSD encounters and will make news paper front pages too.

Quite a difference.
- Lucid uses next gen cameras, Tesla uses last gen cameras.
- Lucid uses next gen forward radar, Tesla uses last gen radar from 2010
- Lucid uses forward Lidar, Tesla doesn't use Lidar.

That alone with okay software should eliminate the issues AP have.

 

[diplomat33]

I think the biggest problem with offering L4 in this current sensor suite they have is compute redundancy and sensor cleaning. Because their current sensor suite is sufficient for L4 highway since its multi modal (360 cameras & 360 radars) and triple modal redundancy in the front.

Thanks. That makes sense.

 

[Soda Popinski]

Quite a difference.
- Lucid uses next gen cameras, Tesla uses last gen cameras.
- Lucid uses next gen forward radar, Tesla uses last gen radar from 2010
- Lucid uses forward Lidar, Tesla doesn't use Lidar.

That alone with okay software should eliminate the issues AP have.

Which FSD computer is Lucid using? I'm assuming EyeQ5?

 

[J1mbo]

To me, it does not make sense to rely on a method that will take longer to solve the problem, if it solves it at all, and prolong the risk of an accident, when there is a better method that will be easier and more reliable.

This mindset is why no-one was building mass-market electric cars or reusuable rockets until Musk came along.

The Lucid sensor suite is overkill for the L3 features they are describing, and no doubt there will be limits on when, where and how L3 can be used (limits in addition to the recent UN mandate)

Maybe "number of cameras" is going to become a marketing thing.

 

[diplomat33]

The Lucid sensor suite is overkill for the L3 features they are describing, and no doubt there will be limits on when, where and how L3 can be used (limits in addition to the recent UN mandate)

I don't think the sensors are overkill for L3. In fact, that sensor suite is pretty standard for safe and reliable L3. The sensor suite will ensure safe and reliable L3. But even if they are overkill, so what? Better to have too many sensors and have safe and reliable L3 then to not have enough sensors and not have reliable L3. What's the point of having less sensors if you don't achieve reliable L3?

Also, the Lucid sensor suite is designed to be future proof. It's designed to give them a robust sensor suite that will be good enough for future features that can be delivered via OTA update. So it makes sense to put more sensors so that you have enough for whatever future L3 features you come up with later. I think the sensor suite is awesome!

 

[Huskyf]

Quite a difference.
- Lucid uses next gen cameras, Tesla uses last gen cameras.
- Lucid uses next gen forward radar, Tesla uses last gen radar from 2010
- Lucid uses forward Lidar, Tesla doesn't use Lidar.

That alone with okay software should eliminate the issues AP have.

Completly wrong about the radar in 2017 Tesla changed the radar is not Bosh anymore....And the new radar have more power distance

 

[J1mbo]

I don't think the sensors are overkill for L3. In fact, that sensor suite is pretty standard for safe and reliable L3. The sensor suite will ensure safe and reliable L3. But even if they are overkill, so what? Better to have too many sensors and have safe and reliable L3 then to not have enough sensors and not have reliable L3. What's the point of having less sensors if you don't achieve reliable L3?

Also, the Lucid sensor suite is designed to be future proof. It's designed to give them a robust sensor suite that will be good enough for future features that can be delivered via OTA update. So it makes sense to put more sensors so that you have enough for whatever future L3 features you come up with later. I think the sensor suite is awesome!

More sensors /= "safe and reliable"

More sensors = more complex software + more potential points of failure

Future features are limited by the APE: EyeQ4 can apparently handle a similar suite of sensors, but the chip is unlikely to be capable of L5 with just OTA updates.

 

[diplomat33]

More sensors /= "safe and reliable"

More sensors = more complex software + more potential points of failure.

That's a very simplistic view. Sure, there is always the potential for failure if you add more sensors but more sensors can also give you the needed data you need to make your system better. If the added sensors help prevent accidents, then it decidedly will make your system safer and more reliable. So you need to find that threshold between simple but not enough sensors and more sensors but too complex.

Just having a simple system with less risk of failure is not necessarily better. For example, having 8 cameras and a front radar and 12 ultrasonics adds complexity and risk of failure, so why doesn't Tesla only have 4 cameras instead? Or better yet, why does Tesla not just do FSD with 1 camera, that would be the simplest system possible with only one point of failure. That would be better no? Of course, we know it would not be because 1 camera is not enough for FSD. So just reducing your sensors to keep things simpler is not automatically the best approach. You still need enough sensors to get the job done reliably!

Future features are limited by the APE: EyeQ4 can apparently handle a similar suite of sensors, but the chip is unlikely to be capable of L5 with just OTA updates.

Uh?! I never said the future features would be L5. Lucid has said that they are not even considering L4 on this sensor suite. So any future features would still be more advanced L3.

 

[J1mbo]

That's a very simplistic view. Sure, there is always the potential for failure if you add more sensors but more sensors can also give you the needed data you need to make your system better. If the added sensors help prevent accidents, then it decidedly will make your system safer and more reliable. So you need to find that threshold between simple but not enough sensors and more sensors but too complex.

Just having a simple system with less risk of failure is not necessarily better. For example, having 8 cameras and a front radar and 12 ultrasonics adds complexity and risk of failure, so why doesn't Tesla only have 4 cameras instead? Or better yet, why does Tesla not just do FSD with 1 camera, that would be the simplest system possible with only one point of failure. That would be better no? Of course, we know it would not be because 1 camera is not enough for FSD. So just reducing your sensors to keep things simpler is not automatically the best approach. You still need enough sensors to get the job done reliably!

Again, you are confusing the number of sensors with system capability. Proof that the marketing works!

There are 360 solutions that only use 1 camera but that approach fails the redundancy & packaging tests. The world isn't ready for cars with sensor domes on the roof just yet.

As we are all AP beta testers, we have seen some of the failures of Tesla's systems, the best example of my point being phantom braking. This is caused by a disagreement between the radar and vision systems. AP is designed to "fail safe", so it brakes until the disagreement is resolved. I understand that Tesla use "whitelists" in their maps as a workaround for this.

The more sensors you introduce, the more overlap you have between them, the higher the probability of disagreement.

Does this make the system safer? You just have to check out the treads about phantom braking to understand how people feel about trusting a nervous ADAS, and how lucky they were that no-one was close enough to rear-end them at the time it happened.

As for Lucid, I imagine they will have to create whitelists for each radar

 

[diplomat33]

Again, you are confusing the number of sensors with system capability. Proof that the marketing works!

I am not confusing anything. Sometimes, you do need more sensors. You need 360 degree coverage with at least 2 sensors so that you have redundancy in case one sensor is wrong.

There are 360 solutions that only use 1 camera but that approach fails the redundancy & packaging tests. The world isn't ready for cars with sensor domes on the roof just yet.

Wait. We are talking about Waymo now? I thought we were talking about Lucid. Lucid does not have a sensor dome on the roof.

As we are all AP beta testers, we have seen some of the failures of Tesla's systems, the best example of my point being phantom braking. This is caused by a disagreement between the radar and vision systems. AP is designed to "fail safe", so it brakes until the disagreement is resolved. I understand that Tesla use "whitelists" in their maps as a workaround for this.

The more sensors you introduce, the more overlap you have between them, the higher the probability of disagreement.

Does this make the system safer? You just have to check out the treads about phantom braking to understand how people feel about trusting a nervous ADAS, and how lucky they were that no-one was close enough to rear-end them at the time it happened.

As for Lucid, I imagine they will have to create whitelists for each radar

And if you just have one sensor and it does not see an object, you will hit something. How is that safer?

I'll gladly take some phantom braking over crashing into stopped vehicles. So yes, if it means that it does not crash head long into stopped vehicles, I think it makes it safer.

 

[WarpedOne]

Phantom braking is just Tesla being incompetent, we all know Lucid will have no phantom braking.
Because reasons ... yada yada yada

Deliver the car, let it be parallel tested => conclusions.

 

[diplomat33]

Phantom braking is just Tesla being incompetent.
We all know Lucid will have no phantom braking. Because reasons ...

I am not saying that Lucid will have no phantom braking at all. But I think it is is silly to say that the answer is to have a sensor suite that is not good enough because if we add the sensors we need, it might create more phantom braking.

I mean we have Teslas that hit large semi trucks in the middle of the road but you guys are like "no, don't add one cheap lidar in the bumper that would solve the problem because we are worried about phantom braking."

 

[WarpedOne]

>> that would solve the problem

You do not know this nor you know what other problems would lidar introduce.

Hitting large semi truck in the middle of the road can be easily avoided by just using the car as it was intended and every law on earth still demands of you - pay attention. If you are not, no lidar will save you.

 

[diplomat33]

>> that would solve the problem

You do not know this nor you know what other problems would lidar introduce.

Yes, I think lidar would help solve the problem because lidar would bounce lasers off the truck and has sufficient resolution tell you that there is an object there. And unlike radar, lidar can be used to detect stationary objects.

Hitting large semi truck in the middle of the road can be easily avoided by just using the car as it was intended and every law on earth still demands of you - pay attention. If you are not, no lidar will save you.

But we are talking about L3 where the driver does not need to pay attention. How do you propose to solve the problem without requiring the driver to pay attention? Yes, lidar with cameras and radar will solve the problem without needing the driver to pay attention. That is why Lucid is adding a front lidar in the first place. They are not adding an extra sensor just for kicks to make their hardware more expensive for nothing.

 

[WarpedOne]

>> How do you propose to solve the problem without requiring the driver to pay attention?

I have no proposal for such a thing and I do not pretend I am capable of teaching those people assigned this task how to solve it.

Know your limits.

 

[diplomat33]

I have no proposal for such a thing and I do not pretend I am capable of teaching those people assigned this task how to solve it.

Know your limits.

But it's not me claiming I know best. The experts who are assigned the task of solving it, say that this is how you solve it. They say it can be solved with cameras, radar and lidar.