Autonomous Car Progress

[stopcrazypp]

• Personal ownership of self-driving cars will be rare, expensive and makes little sense.
  • Especially because of the required super-sensor-hardware
  • This will have cultural impact in several ways, including discussion of teenage experiences and relationships
• The required remaining improvements (the last fractional-percent of use-cases) are very hard, will take time, will make vehicles expensive

This part stood out to me. A rep from Waymo is noticeably missing, but this panel represents most of the major L4/L5 players and it seems they are all of the opinion L4/L5 would be too expensive for personal ownership.

This had always been an arguing point in previous discussion. The idea is pushed that LIDAR costs would drop so much that it'll be viable for personal owned vehicles and then that would be the pathway to affordable L4/L5. That kind of handwaves away LIDAR sensors can have way different capabilities (from the earliest simple ones used for cruise control, to the most expensive 360 degree rotating ones). The solution for the more sanely priced cars (like the $102k Honda, which is still quite expensive and likely Honda loses money on with only 100 units lease-only) uses 5 units with less FOV (not the rotating ones). Might be while before we see it be affordable enough to we can see in a $40k car.

 

[Doggydogworld]

For L3, it was more than just "maybe unrealistic". The guy said that L3 doesn't work and was only included to make the SAE nomenclature look more continuous. He explained that it takes about 30 seconds for a human to fully re-engage again with the road after not paying attention for awhile....

That's an opinion. Those who say 10 seconds are no less credible. It's tough to do realistic studies, so there's a lot of "coming up with the desired answer".

.... but avoiding accidents often requires split second intervention.

You don't intervene to avoid accidents with Level 3. That's FSD Beta. You take over with Level 3 when it approaches the freeway exit, or when the traffic jam ends , or when the blizzard gets too severe and system confidence degrades. Some pedants say this means Level 3 is the same as Level 4, since both handle all driving tasks within their ODD.

I'm not sold on the personal ownership thing. When it comes to cars people pay a lot of money for stuff that doesn't make economic sense. I think most will resist giving up their cars, except in cities where parking is an expensive hassle. We'll see.

 

[Bladerskb]

This part stood out to me. A rep from Waymo is noticeably missing, but this panel represents most of the major L4/L5 players and it seems they are all of the opinion L4/L5 would be too expensive for personal ownership.

They are robot taxi companies. Its a light switch for them, there's no in-between. They can't release anything until its 100% working. However companies like Mobileye and Huawei are innovating from the other end.

This had always been an arguing point in previous discussion. The idea is pushed that LIDAR costs would drop so much that it'll be viable for personal owned vehicles and then that would be the pathway to affordable L4/L5. That kind of handwaves away LIDAR sensors can have way different capabilities (from the earliest simple ones used for cruise control, to the most expensive 360 degree rotating ones). The solution for the more sanely priced cars (like the $102k Honda, which is still quite expensive and likely Honda loses money on with only 100 units lease-only) uses 5 units with less FOV (not the rotating ones). Might be while before we see it be affordable enough to we can see in a $40k car.

I'm lost, why are you extrapolating from the price of a single car that has nothing to do with the price of Lidar?
Also why are you settling at $40k. If I wanted to buy a Model 3 "pure vision" FSD, I would be paying +$50k?

There's not a-lot of info on Honda's Lidar but it looks like their two forward lidars are Valeo Scalar Generation 2 lidars. I would be shocked if they cost more than $250 dollars.

If we look at just 2021 there are multiple cars releasing with lidars. None of them come close to $104k. You have Xpeng P5 with 2 high resolution lidars that cost ~$35k, Huawei's Arcfox Alpha S Hi version has 3 high resolution lidars and cost $60,000 - $66,000 (max). The hardware on Arcfox Alpha S Hi version is orders of magnitude more complex than is in the Honda car.

OEMs try to get very high margins on their return for ADAS. GM's Supercruise 1.0 used EyeQ3 that cost around $30 with a single camera ($10-15) and single front radar (around $50-$70) and sells it for $2,500 but its locked to high trims. The 2.0 has more radars and uses $60 EyeQ4.

Now for the actual figures: Super Cruise is listed as a $2,500 option on the 2021 Cadillac CT5 online configurator, but the actual price change when opting for the feature is $8,740. Add in the costs of the V6 engine, AWD, and the various necessary packages, and the total price jumps to $58,470. That is the lowest possible price for a 2021 CT5 fitted with Super Cruise, but it’s still $16,680 more than the MSRP of a base Premium Luxury model. The other way to configure Super Cruise on the CT5 Premium Luxury is to choose the optional Platinum Package, in which case the price rises to $61,640, or $19,850 above MSRP of the base model.

For a BOM of around $100-$150 that's a ridiculous multiplier.

 

[stopcrazypp]

That's an opinion. Those who say 10 seconds are no less credible. It's tough to do realistic studies, so there's a lot of "coming up with the desired answer".

Audi's system gave 10 seconds. Audi has since ditched the L3 idea completely.
2019 Audi A8 Level 3 autonomy first-drive: Chasing the perfect ‘jam’

A redditor claims the Honda system allows 30 seconds to take over, but I looked through all the links and none of them say how many seconds is allowed. I only found articles that mention a "few" seconds. The CGI demo video at around 20:30 showed the seat belts start tensioning and preparing to stop the car after around 10 seconds of no response (and alarm turns from orange to red), and the car starts pulling over at around 20 seconds. Doesn't seem like you really have 30 seconds for a handover.

https://www.reddit.com/r/SelfDrivingCars/comments/lyh0ra

You don't intervene to avoid accidents with Level 3. That's FSD Beta. You take over with Level 3 when it approaches the freeway exit, or when the traffic jam ends , or when the blizzard gets too severe and system confidence degrades. Some pedants say this means Level 3 is the same as Level 4, since both handle all driving tasks within their ODD.

Is not just that. You also have to take over when the system requests you to do so. That's the main difference between L3 and L4. With L4 you never have to respond to the system even if it requests you to. With L3, you must respond, or it can't guarantee you won't get into an accident. All L3 does is give you an extra few seconds to respond (while L2 gives you no seconds to respond, you must be ready at all times).

I'm not sold on the personal ownership thing. When it comes to cars people pay a lot of money for stuff that doesn't make economic sense. I think most will resist giving up their cars, except in cities where parking is an expensive hassle. We'll see.

Totally depends on how the government handles things. If the government decides it's much safer to only let self driving cars operate, we may have laws that ban cars that are driven by drivers. Of course in a democratic government, this all depends on the opinion of the overall population, but that's not true of all governments. There may also be a demographic change where a new generation have grown up used to self driving cars and may never see the need to drive themselves (kind of like the current generation growing up never having to drive a manual transmission vehicle).

 

[gearchruncher]

kind of like the current generation growing up never having to drive a manual transmission vehicle

I know this is a fun "Millennial anti-theft device" meme, but the last time 50% of cars sold in the USA had manual transmissions was in the 70's. In 1980 2/3rds were automatic. Nobody under about 55 needed to learn to drive a manual, so it's been about the last 3-4 generations that haven't needed to.

It's going to be a very long time until 50% of the vehicles on the road are self driving. Even if it was technically feasible today, you're looking at 10-20 years due to economics.

 

[stopcrazypp]

I know this is a fun "Millennial anti-theft device" meme, but the last time 50% of cars sold in the USA had manual transmissions was in the 70's. In 1980 2/3rds were automatic. Nobody under about 55 needed to learn to drive a manual, so it's been about the last 3-4 generations that haven't needed to.

It's going to be a very long time until 50% of the vehicles on the road are self driving. Even if it was technically feasible today, you're looking at 10-20 years due to economics.

Yeah, it might take a while, but current market share of stick shift is around 2%. It may be well before this kind of share when you have the majority of the population that might not see the point of a car you have to manually drive.
https://www.carmax.com/articles/stick-shift-index

 

[gearchruncher]

Yeah, it might take a while, but current market share of stick shift is around 2%. It may be well before this kind of share when you have the majority of the population that might not see the point of a car you have to manually drive.

I think you missed my point. Plenty of people that learned to drive in 1980 never needed to learn to drive a manual. You don't need to wait until only 2% of cars are manuals- it's pretty easy to get through your life never driving one when 33% or less are manuals. It's far from just people born in 2000 that never learned. It took from ~1970 to ~2020 (50 years) for the manual to go from 50% to 2%. Hence my point that while it is inevitable that eventually we have people that never thought of cars as anything but self driving, that is really far away.

I don't think that will mean much for people not considering not owning a car. Plenty of people in dense cities like NYC that have been able to hire transportation more affordably than owning (Cab, then Uber/Lyft) have been doing this for decades. The instant the service is better and the cost cheaper, huge percentages of Americans won't care about ownership at all.

 

[stopcrazypp]

They are robot taxi companies. Its a light switch for them, there's no in-between. They can't release anything until its 100% working. However companies like Mobileye and Huawei are innovating from the other end.



I'm lost, why are you extrapolating from the price of a single car that has nothing to do with the price of Lidar?
Also why are you settling at $40k. If I wanted to buy a Model 3 "pure vision" FSD, I would be paying +$50k?

Point is you get the car with the hardware installed for $40k and the automaker is still making a margin on it at that price. Adding more sensors will likely drive the break even price significantly higher and the price to the end consumer typically is not 1:1 anyways (it's a multiplier of the BOM).

There's not a-lot of info on Honda's Lidar but it looks like their two forward lidars are Valeo Scalar Generation 2 lidars. I would be shocked if they cost more than $250 dollars.

It's unlikely for it to be that cheap given it's priced at $600 / unit at high volume and 500 units is not high volume.

Money For Everythin’

Will pure-play LiDAR companies succeed in disrupting the price-performance equation as LiDAR becomes an enabling technology for ADAS and AVs? Or will the automotive Tier 1s continue to dominate as transportation transforms into a more automated and safe environment?

www.forbes.com

If we look at just 2021 there are multiple cars releasing with lidars. None of them come close to $104k. You have Xpeng P5 with 2 high resolution lidars that cost ~$35k, Huawei's Arcfox Alpha S Hi version has 3 high resolution lidars and cost $60,000 - $66,000 (max). The hardware on Arcfox Alpha S Hi version is orders of magnitude more complex than is in the Honda car.

Well, let's see how much those cars cost first with sensors included (not just base price, which may not necessarily include them, Honda for example is not including those sensors in the base model).

OEMs try to get very high margins on their return for ADAS. GM's Supercruise 1.0 used EyeQ3 that cost around $30 with a single camera ($10-15) and single front radar (around $50-$70) and sells it for $2,500 but its locked to high trims. The 2.0 has more radars and uses $60 EyeQ4.

For a BOM of around $100-$150 that's a ridiculous multiplier.

That's my point also, the OEMs will charge much multiplier than it costs them for BOM. 5 sensors that cost about $3000 total (lets assume the future high volume price) will likely cost a lot more to the consumer. Maybe the Chinese manufacturers will play a different game, but it may be a while before we even see a single one of those cars hit US shores.

 

[mark95476]

Cost aside, I'd like to see an OEM try selling a car with a huge LIDAR sensor(s) blob to an individual. Something like that will only sell to commercial services and fleet owners.

Point is you get the car with the hardware installed for $40k and the automaker is still making a margin on it at that price. Adding more sensors will likely drive the break even price significantly higher and the price to the end consumer typically is not 1:1 anyways (it's a multiplier of the BOM).

It's unlikely for it to be that cheap given it's priced at $600 / unit at high volume and 500 units is not high volume.

Money For Everythin’

Will pure-play LiDAR companies succeed in disrupting the price-performance equation as LiDAR becomes an enabling technology for ADAS and AVs? Or will the automotive Tier 1s continue to dominate as transportation transforms into a more automated and safe environment?

www.forbes.com

Well, let's see how much those cars cost first with sensors included (not just base price, which may not necessarily include them, Honda for example is not including those sensors in the base model).

That's my point also, the OEMs will charge much multiplier than it costs them for BOM. 5 sensors that cost about $3000 total (lets assume the future high volume price) will likely cost a lot more to the consumer. Maybe the Chinese manufacturers will play a different game, but it may be a while before we even see a single one of those cars hit US shores.

 

[Doggydogworld]

Cost aside, I'd like to see an OEM try selling a car with a huge LIDAR sensor(s) blob to an individual. Something like that will only sell to commercial services and fleet owners.

This Volvo design (not shipping) is the only one I've seen with a lidar "bulge". All shipping cars hide their lidars in the grill, behind the glass and/or in headlight/taillight assemblies.

 

[Bladerskb]

Point is you get the car with the hardware installed for $40k and the automaker is still making a margin on it at that price. Adding more sensors will likely drive the break even price significantly higher and the price to the end consumer typically is not 1:1 anyways (it's a multiplier of the BOM).

That doesn't matter especially when we are talking about personally owned robot-taxi. If waymo were selling their car with their system today for 35k, it doesn't matter what the car cost without their system.

You pay $35k and you get the entire system (hardware and software).

In the same way the current costs are:

Tesla's Model S system (8 cameras and one radar, 144 Tops Tesla computer) at the moment cost +$80k.

Nio's ET7 Model S competitor (11 cameras, 5 radars, 1 Hi-res lidars, 1000 Tops Nvidia Computer) at the moment cost ~$69k.

Huawei's Arcfox Alpha S system (3 Hi-res Lidars, 12 cameras, 6 radars & 400 TOPS Huawei 400 TOPs computer) at the moment cost $60-66k.

Tesla's Model 3 system (8 cameras and one radar, 144 Tops Tesla computer) at the moment cost $50k.

Xpeng's P5 system (12 cameras, 6 radars, 2 Hi-res lidars, ? Tops Nvidia computer) at the moment cost $35k.

It's unlikely for it to be that cheap given it's priced at $600 / unit at high volume and 500 units is not high volume.

Money For Everythin’

Will pure-play LiDAR companies succeed in disrupting the price-performance equation as LiDAR becomes an enabling technology for ADAS and AVs? Or will the automotive Tier 1s continue to dominate as transportation transforms into a more automated and safe environment?

www.forbes.com

Well, let's see how much those cars cost first with sensors included (not just base price, which may not necessarily include them, Honda for example is not including those sensors in the base model).

That's my point also, the OEMs will charge much multiplier than it costs them for BOM. 5 sensors that cost about $3000 total (lets assume the future high volume price) will likely cost a lot more to the consumer. Maybe the Chinese manufacturers will play a different game, but it may be a while before we even see a single one of those cars hit US shores.

Interesting because they have such low resolution compared to others. But Its not surprising that their only partners are traditional automakers who probably spec'ed this out 3-4 years ago and are allergic to change so they are stuck with it. Other OEMs lidars that have orders of magnitude more resolution for less money. Luminar who has one of the best resolution on the automotive market and cost $500. Others like robosense that cost $200, Huawei is trying to get their lidar down to $200 which means they are hovering around that number. Livox is most likely around $100-$250 aswell. Innoviz one is just under $1,000.

To put that into perceptive, these lidars used to cost around 75k-100k. So we are talking about orders of magnitude improvement on cost and quality

 

[Bladerskb]

This Volvo design (not shipping) is the only one I've seen with a lidar "bulge". All shipping cars hide their lidars in the grill, behind the glass and/or in headlight/taillight assemblies.

Nio ET7

 

[diplomat33]

Argo AI, the autonomous driving startup backed by Ford and VW, has revealed a new long-range lidar. The company's CEO Bryan Salesky says it gives Argo an edge in deploying self-driving cars commercially. Salesky spoke exclusively to Bloomberg's Ed Ludlow about the technology, its financing plans and regulation.

 

[S4WRXTTCS]

this panel represents most of the major L4/L5 players and it seems they are all of the opinion L4/L5 would be too expensive for personal ownership.

To really benefit from self-driving cars we need to be able to upgrade them on a regular basis to take advantage of new standards, and new technologies.

To me this is where most of the expense will come in.

I also question whether ownership model really makes any sense anymore for the majority. Having a vehicle demands a lot of space, and commitment.

You have to have a place to park it both at home, and at your destination
You have to carry insurance (assuming its a mixed mode vehicle)
You have to wash it
You have to maintain it
You have to charge it

Even the Honda L3 vehicles in Japan don't allow the ownership. As I understand it they're just leases.

 

[gearchruncher]

You have to carry insurance (assuming its a mixed mode vehicle)

The current laws in some states say that a self diving car must be insured, and any violations or liability are the responsibility of the registered owner of the vehicle. Who would want to own one with those rules? The algorithm designer uploads a change to your car and you're fully liable?

 

[rxlawdude]

The current laws in some states say that a self diving car must be insured, and any violations or liability are the responsibility of the registered owner of the vehicle. Who would want to own one with those rules? The algorithm designer uploads a change to your car and you're fully liable?

Now you're a lawyer?

 

[DanCar]

The current laws in some states say that a self diving car must be insured, and any violations or liability are the responsibility of the registered owner of the vehicle. ...

I wonder which states are those.

 

[gearchruncher]

I wonder which states are those.

North Carolina: NC Regulates Fully Autonomous Vehicles (summary)
Actual bill: https://www.ncleg.net/Sessions/2017/Bills/House/PDF/H469v7.pdf

 

[gearchruncher]

Now you're a lawyer?

Seriously? In one thread you say everyone should read their contracts and that makes it clear that Tesla doesn't have to deliver you anything, and then in another your suggest you have to be a lawyer to understand laws? Which one is it?

I think the average non-lawyer American can understand "The person in whose name the fully autonomous vehicle is registered is responsible for a violation of this Chapter that is considered a moving violation, if the violation involves a fully autonomous vehicle. " well enough to decide if they are willing to own a self driving vehicle and let it drive around and any laws it breaks are the responsibility of the owner, not the manufacturer, and that laws like this might discourage individual ownership.

 

[orion2001]

I also question whether ownership model really makes any sense anymore for the majority. Having a vehicle demands a lot of space, and commitment.

You have to have a place to park it both at home, and at your destination
You have to carry insurance (assuming its a mixed mode vehicle)
You have to wash it
You have to maintain it
You have to charge it

Even the Honda L3 vehicles in Japan don't allow the ownership. As I understand it they're just leases.

I just don't buy the robotaxi replacing personally owned cars argument. I get the conveniences, and I get their value in a dense urban environment where car ownership is already a luxury and a hassle due to limited space for parking and higher costs of ownership. But, the huge thing being overlooked here is that you have to bring along everything you need at all times with you when using a robotaxi because you can never store things in one because it isn't yours.

For families with kids and/or pets, that already becomes a pretty terrible value. Lug around car seats and install them every time you get in a robotaxi. That's just never going to fly. I always have a dog crate in our minivan along with some hiking supplies, hiking sticks, etc, so that going on hikes on the weekend with my kids and dog is just a matter of getting all of us into our car and heading over to our favorite hiking spots. With a robotaxi, you gotta lug all that stuff along, get it into the taxi, then go wherever... and then what? You pay to have the car wait for you with your crate and car seats in the parking lot for several hours while you hike, then return home and remove all of that again?

I'm just not buying the argument that robotaxis will ever replace ownership of vehicles. It will make sense for a subset of society, but I just don't see anyway for it to be the only (or even dominant) mode of transportation.