[likeXC90]

MobilEye uses the fully camera-based "Supervision" in hardware available now with the potential to add Lie-dar later.

The fact Luminar Lie-dar - or any Lie-dar for that matter - has not been chosen for the initial Polestar 4 release coupled with Volvo CEO's "corrected" statement that (Luminar) Lie-dar is not the reason for the EX90 delay, doesn't quite build confidence in the integration (and/or performance?) of Lie-dar into current products (ex: EX90).

"Lie-dar - just not today!" [emoji12]

Polestar to work with Mobileye on autonomous driving technology in Polestar 4

Polestar and Mobileye are to cooperate to make Polestar 4 the first production car to feature Mobileye’s Chauffeur autonomous driving technology – to be integrated by ECARX and with potential application in future Polestar vehicles. Mobileye produces leading advanced driver assistance system...

media.polestar.com

Polestar 4 to feature Mobileye's hands-off, eyes-off driving tech | TechCrunch

Swedish electric vehicle maker Polestar wants to bring Mobileye's Chauffeur, a hands-off, eyes-off autonomous driving technology, to its Polestar 4

techcrunch.com

 

[medels]

MobilEye uses the fully camera-based "Supervision" in hardware available now with the potential to add Lie-dar later.

That 100% part of Geely hardware already found on Zeekr product.
the Zeekr 001 phase 2 / Lotus Eltre and all premium brand using SEA will be soon offered with a top mounted Lidar and a US qualcomm chip


Volvo is loosing competitiveness within the group as their own solution with Nvidia and Luminar is hugely delayed.

 

[rfkuehn]

The fact Luminar Lie-dar - or any Lie-dar for that matter - has not been chosen for the initial Polestar 4 release coupled with Volvo CEO's "corrected" statement that (Luminar) Lie-dar is not the reason for the EX90 delay, doesn't quite build confidence in the integration (and/or performance?) of Lie-dar into current products (ex: EX90).

Lie-dar or LieDAR?

the Zeekr 001 phase 2 / Lotus Eltre and all premium brand using SEA will be soon offered with a top mounted Lidar and a US qualcomm chip

Volvo is loosing competitiveness within the group as their own solution with Nvidia and Luminar is hugely delayed.

P*3 w/out LiDAR has also been delayed, not just w/ LiDAR - so are these delays beyond just problems w/ LiDAR itself? And/or with integration w/ other SW?

SPA2 strikes me as an absolute mess, EX90 (fka Emblah) was originally due in 2021 for MY22 as prior CEO Samuelsson's crowning achievement before retiring March 2022. Now tracking to three years late and counting.

I wonder what is becoming of V546 EX70-whatever - going on SPA2 as direct counterpart to P*3, or to SEA/GPA instead? SPA2 seems like a bad bet, but dealers have seen it twice, right? Miami 2022 + Charleston 2023? 5x EVs - EX90, 30, 60, 70, plus ES90? Too late to change the production plan, whatever it is? Perhaps V546 on SPA2 has been slid to the right in favor of non-SPA2 cars, then gets dusted off once more progress is made w/ SPA2 delays?

 

[VolvoGoteborg]

Why are you calling it Lie-dar?

 

[VolvoGoteborg]

That 100% part of Geely hardware already found on Zeekr product.
the Zeekr 001 phase 2 / Lotus Eltre and all premium brand using SEA will be soon offered with a top mounted Lidar and a US qualcomm chip


Volvo is loosing competitiveness within the group as their own solution with Nvidia and Luminar is hugely delayed.

So SEA platform has adopted Mobileye system ADAS to later be supplemented with lidar for autonomous drive mode using qualcomm chip.

Volvo's Nvidia and Luminar system is more in-house than the Mobileye route taken by Geely? and inherited by Lotus and Polestar for the (seemingly) Zeekr 001 derived Eltre and 4 respectively. Albeit, Zeekrs are also designed in Sweden.

More in-house is better I think but a shame it is so delayed. Would end up on the three SPA-2 cars i.e. Ex90, Polestar 3 and "ES90". The V546 is also SPA-2? Anything SEA seemingly would adopt the Mobileye solution.

As I've said before the delays to SPA-2 seem to have caused otherwise originally planned for SPA-2 cars (so 60 and 90 series replacements albeit one sedan and one wagon) to switch over to the SEA = GPA platform instead.

Volvo also seems to be going alone with its own central computer hardware and software (rather than hardware from suppliers, modules with base software from suppliers which the communicate on a Volvo car operating system. As per Volvo's Tech Day info. Presumably all planned for SPA-2 cars (as above) but will continue for SEA = GPA Volvos presumably rather than adopt Geely's setup for SEA platform and Geely electrical system (which Polestar 4 also uses).

CMA cars (XC40 and C40) use a different system to SPA-1 cars?

EX30 seems standalone or Geely electrical system as Volvo's SPA-2 one isn't yet ready.

EM90 will be Geely electrical system too as SPA-2 one also isn't ready.

Getting a little messy but should settle down to Volvo's system for SPA-2 and SEA for Volvo which is GPA.

I prefer the idea of Volvo doing their own thing with their own designed software and hardware and Volvo OS for SPA-2 and SEA = GPA. But objectively for economies of scale for hardware and software testing it would make more sense for Geely's electrical system in SEA to be rolled out as far and as wide as possible. Given the time, cost and complexity it is now taking to develop truely robust software systems. See VWs suffering with this an example of how it can go very wrong.

 

[rfkuehn]

Why are you calling it Lie-dar?

I think his point was re: seemingly endless delays, Lie-dar meaning its kind of like a vapor offer that hasn't materialized.

I then offered "LieDAR" in the spirit of continuous improvement.

It won't come to XC90 refresh I'm waiting for, so really just watching how it plays out.

 

[VolvoGoteborg]

That 100% part of Geely hardware already found on Zeekr product.
the Zeekr 001 phase 2 / Lotus Eltre and all premium brand using SEA will be soon offered with a top mounted Lidar and a US qualcomm chip


Volvo is loosing competitiveness within the group as their own solution with Nvidia and Luminar is hugely delayed.

Interesting pic there medels. What's that from?

 

[rfkuehn]

Will deployable LiDAR eventually become the standard expectation?

 

[medels]

Interesting pic there medels. What's that from?

Soon to be release Zeekr sedan.

 

[VolvoGoteborg]

Actually, further to what I said above, I suppose that just as Volvo will be using their own hardware and software in SPA-2 cars as well as SEA = GPA cars, I suppose they can too also roll out their out Lidar + Nvidia autonomous mode as well instead of the Geely group adopted Mobileye chauffeur setup. Maybe (and would make more sense) for Geely to then transition to and adopt the in-house, originally led by Volvo) Nvidia based system. If one assumes that in-house is better and more profitable than buying things off the shelf from Tier 1 suppliers.

 

[likeXC90]

Lie-dar or LieDAR?

Why not both? 😜

SPA2 strikes me as an absolute mess, EX90 (fka Emblah) was originally due in 2021 . . . Now tracking to three years late and counting.

Agreed. Like you suggested, software seems to be the common underlying theme here. And Lie-dar is extreeeeemely software intensive, on top of everything else new (ie the car's OS and nearly all electric systems). I don't know how they can make development even harder with so many technologies new to Volvo, at least, and maybe new also to their suppliers as well, as everyone is rushing - literally falling over themselves - to be the first to market with their latest whiz-bang gizmo car (ie, share of the market).

Why are you calling it Lie-dar?

I have worked with several Lie-dar startups from the beginning bubble in the mid-2010s and have yet to see the promise of "cold fusion" take place at the mass production level in more "mature" Lie-dar manufacturers. There are so many technical development and commercial challenges in and of itself (forget the regulatory, infrastructure, legal, etc challenges that Lie-dar unleashes) that the former lofty goals have now been reduced to something more realistic - it's no longer autonomous driving but now Level 3 or Level 2+. Why pay thousands more for Level 3 when competing technolgies can do Level 3 for cheaper with existing technology? I have no doubt Lie-dar for automotive will happen but - "just not today."

Actually, further to what I said above, I suppose that just as Volvo will be using their own hardware and software in SPA-2 cars as well as SEA = GPA cars, I suppose they can too also roll out their out Lidar + Nvidia autonomous mode as well instead of the Geely group adopted Mobileye chauffeur setup. Maybe (and would make more sense) for Geely to then transition to and adopt the in-house, originally led by Volvo) Nvidia based system. If one assumes that in-house is better and more profitable than buying things off the shelf from Tier 1 suppliers.

I believe Volvo made a mistake - they have admitted to this in bits and pieces but seem to avoid saying it all together outright - trying to simultaneously develop into one vehicle all these tasks in parallel:

• a new platform (SPA2)
• a new car (EX90)
• a new all-electric car (not based on a previous gas model, like the XC40)
• a new car OS (Google)
• a new, highly intensive Lie-dar sensor-based suite
• and what else? (new manufacturing methods? new suppliers? etc)

Traditionally, new technology is incrementally tried and debugged to reduce risk but lumping "all the eggs together in one basket" is what the EX90 sounds like. Recycling MobilEye from Geely into Polestar 4 sounds more in line with traditional development and, perhaps now, a "lessons-learned" approach.

I do want Volvo to succeed - I'm looking forward to getting a Volvo full EV someday - but please, clean up this mess and clean it well. My 2 cents . . .

(It makes me wonder if Samuelsson and Green were ousted for moving too fast or not fast enough towards full EV - though I'm betting not fast enough!)

 

[rfkuehn]

I believe Volvo made a mistake - they have admitted to this in bits and pieces but seem to avoid saying it all together outright - trying to simultaneously develop into one vehicle all these tasks in parallel:

• a new platform (SPA2)
• a new car (EX90)
• a new all-electric car (not based on a previous gas model, like the XC40)
• a new car OS (Google)
• a new, highly intensive Lie-dar sensor-based suite
• and what else? (new manufacturing methods? new suppliers? etc)

• a new production line with a newly-expanded workforce within a severely-underutilized, relatively-new factory (South Carolina)

 

[VolvoGoteborg]

Thanks for your insight. Not sure if this is the post or not for this as it's a bit of a parallel topic but related. I'm keen to get some insight from someone working in ADAS systems and autonomous drive. Trouble is most of those who are knowledgeable from working on such things still work on such things and are bogged down in NDAs so can't speak up.

There is a lot of noise from the Tesla-clan / cult / stock pumping enthusiasts on Twitter about the, apparently, revolutionary V12 of "FSD". I'm keen to cut through the crap, the PR, the enthusiasm of fans who don't know what they are looking at and worship every word of Musk and Tesla regardless. I know from others on Twitter who are not fans that 1) US regulation for adaptive cruise and self drive etc is very minimally regulated. No certification or validation for example before sale. Will link and article here in due course. More of a "let them launch it, monitor (somehow/badly) and make OEMs recall it if there are problems. EU is the opposite and there's regulation and approval to go through first. I know that "FSD" is overhyped and oversold as "autopilot" and "full self driving" when it really is a Level 2 system that needs constant supervision. But, is v12 really anything significant? AI based apparently whereas all FSD before it is not. Misunderstandings from fans but doesn't seem all the cars using v12 would actually upload shedloads of data base to Tesla HQ to help the system evolve. What data would it be sharing anyway? Data overload. How is it usefully processed etc. This implies each car using v12 learns independently so you end up with a multitude of versions as each car has experienced different things. To me this doesn't sound like a good thing. Not does it sound good that Tesla FSD v12 learns from cars around it. That seems a disaster waiting to happen as it would just learn from endless bad habits of other drivers. Car needs to obey the highway code even if no one around it does so! How would such individual systems share their learning with HQ and other vehicles? Parallel a little I suppose to adaptive automatic gearbox control that learns for each gearbox and driver it is coupled to. In a way then every P2 or P3 Volvo auto is running a different "brain" for the gearbox as they have all adapted independently from the original base software. But a far less risky outcome.

At the same time, Tesla fan boys talk of this new Tesla datacentre which is training FSD v12 using videos.

So.. I know there's a lot of fan / cult / stock pumping / obsessives out of their depth and singing the praises of anything Tesla does. I don't know any other OEM that has such a numerous, active and engaged fan base. Viscous circle of pumping the brand image. Musk owning Twitter also doesn't help other OEMs here. Unsurprisingly Volvo hasn't tweeted since Musk took over. Other Geely brands are still posting as normal. No one has really pushed onto Threads yet (I wish they would!!).

I know that other OEMs and Tier 1s don't have this sort of fan base nor do they push PR of every little thing on social media. There's A LOT going on at other OEMs and Tier 1 suppliers that is kept quiet. Plus lots of other little companies doing clever things too include a number training autonomous drive systems using video examples. Nothing special by Tesla. Just excessive noise and PR.

So... what I'm really keen to know is... is Tesla's FSD V12 and their data centre really something of any significance or just the usual Musk and Tesla fan hyperbole and BS?

For example, in a press release from Volvo (Volvo Cars deepens collaboration with NVIDIA; next-generation self-driving Volvos powered by NVIDIA DRIVE Orin - Volvo Cars Global Media Newsroom ) there's a collaboration for Nvidia AI for autonomous driving. So Volvo quietly developing something very clever. Ok, so SPA-2 software is bogging things down but they are on an innovative path.

So Tesla FSD v12 apparently being AI is not a big deal. Plenty of others working on it? Volvo too! Is Volvo further ahead than other OEMs? I hope so.

Maybe Tesla has an advantage with their massive datacentre to train their FSD v12. Maybe that's an advantage vs other OEMs including Volvo. Maybe Volvo has such a capability but is keeping it quiet. Geely has their new ECARX vehicle electronics dept so maybe they too are working on something similar but keeping it quiet and aren't doing endless social media PR to pump stock prices.

As above, overall it seems Volvo is probably in a good position with Nvidia AI based autonomous drive software development coupled to Lidar. Better in house (better profit margins? "vertical integration" than buying a system off the shelf from a Tier 1 even if that has been the usual way of doing things. Teslas approach may be advantageous or may be deeply flawed.

Would love to get an expert in ADAS and autonomous drive to speak up and give a view. I'd love to hear from many many more. Cameras only for Tesla is cheaper but has been discussed before by experts as a flawed approach but pushed through by Musk. Hopefully the robust EU approval process will highlight issues and bring home to roost Tesla's deep error in taking this approach.

Soz for the long post! Been thinking about this for a while

 

[rfkuehn]

Polestar 4: U.S.-spec model information to be released in 2024

The Polestar 4 will be the company's fourth and potentially one of its best-selling models worldwide. Know all about the Porsche Macan EV rival here.

topelectricsuv.com

 

[FusionRedXC60]

LiDAR:
The best advantage is to detect distance of stationary objects in distance so enough time is allowed for the car to slow down in advance. And LiDAR can be used like a low resolution camera. Weak part is not so good to use image contrast to detect road side curb, objects etc.

Camera:
maybe processing algorithm is more advanced as explored more in-depth by researchers.
However distance detection is the worst part. I guess it has to rely on the fact that object appears bigger (on picture) when you getting closer. It is very rough calculation at best, for distant object especially. Not to mention it has to use a lot of guess to figure out if the object in front is a real vehicle or picture on billboard. This is why early Tesla couldn't recognize stopped car in distance on road. Not sure how it is improved now or not. Maybe researchers in Tesla could make it work to some degree. But fundamentally it is very inaccurate and even flawed.

So I only trust LiDAR as it provides accurate distance data that is much easier to be processed by computer and get more accurate result. This is very important for safety.

 

[VolvoGoteborg]

LiDAR:
The best advantage is to detect distance of stationary objects in distance so enough time is allowed for the car to slow down in advance. And LiDAR can be used like a low resolution camera. Weak part is not so good to use image contrast to detect road side curb, objects etc.

Camera:
maybe processing algorithm is more advanced as explored more in-depth by researchers.
However distance detection is the worst part. I guess it has to rely on the fact that object appears bigger (on picture) when you getting closer. It is very rough calculation at best, for distant object especially. Not to mention it has to use a lot of guess to figure out if the object in front is a real vehicle or picture on billboard. This is why early Tesla couldn't recognize stopped car in distance on road. Not sure how it is improved now or not. Maybe researchers in Tesla could make it work to some degree. But fundamentally it is very inaccurate and even flawed.

So I only trust LiDAR as it provides accurate distance data that is much easier to be processed by computer and get more accurate result. This is very important for safety.

Thank you. That's very helpful.

So nothing changes with regards the Tesla situation. Camera only and a flawed system that is being overhyped. Bubble should burst soon!

 

[jkk]

Thank you. That's very helpful.

So nothing changes with regards the Tesla situation. Camera only and a flawed system that is being overhyped. Bubble should burst soon!

What is not overhyped with regards to Tesla however, is the architecture that Volvo is trying to replicate with SPA2 / central computing. See for example this short part where Ford CEO is talking about the pain points highlighting why this new way of doing this is actually a good idea:

Not too surprising, the project (SPA2) does seem to have a lot of pain points still.

 

[VolvoGoteborg]

Volvo's Tech Day was a good summary of the situation. I've seen this too at the OEMs I work for. I'm not an IT or electrical person (I'm an engine guy) but my understanding is that the status quo is that OEMs buy modules with base software from e.g. Motorolla, Continental, Bosch etc. There's a degree of self development vs complete off the shelf part subtlety between different OEMs. But in general there's an off the shelf module or a partly off the shelf module for all the discrete modules around the car that communicate via CAN or it's successor. The CAN network operating system is e.g. made by the OEM (usually). The bottleneck here is when it takes an age to get to the bottom of if a bug or issue is the OEMs operating system issue or the extra software added to the Tier 1 supplied module by the OEM or an issue with the base software of a module. Everyone does it this way.

Either seeing an opportunity or seeing that Tesla (maybe) took a different approach (just as an aside, Scania and Caterpillar designed and design their own engine ECUs hardware and software in house. Everyone else buys it from Delphi, Bosch or Denso etc) to shift from many discrete modules to a single core computer doing it all. Volvo is going this way. Downside is it's a big task. Upside is you design the hardware and software in house. No Tier 1 base software bottleneck in development or resolving bugs. This route would still have Tier 1s making the modules physically I assume rather than OEMs or Volvo sets up electronic unit production in house.

Geely is for example doing more of this approach via discrete modules with the ECARX division (curiously for me in the UK this company/department has an HQ in London). Geely also moving into making their own chips.

I'm not sure if Tesla doing this (with potential profit margin improvements?) triggered Volvo and others to follow suit or was this parallel independent evolution to take the opportunity to be more self sufficient. Don't know.

 

[VolvoGoteborg]

As another aside the adoption of large part castings (I don't like how Tesla-isms such as gigafactory for batteries, or megacasting or frunk get used by journalists and some OEMs. Gives Tesla too much credit).

I'm not sure if other OEMs are following suit in adopting this as it lowers the production time (and hence cost) per unit you build and they need to do this Tesla type approach to gain those cost and production efficiencies even though objectively it isn't the best thing to do necessarily. Everyone needs to do it to gain the some cost and build time savings. Might it be the best idea though?

Counterpoints are the risk of such large castings e.g. Tesla is having big problems with them cracking which writes off the car. Castings can't be repaired apparently. Insurance costs are higher as any damage to the casting can't be repaired and the car is written off too.

At the moment I see it as being adopted as a need to improve build cost and speed but isn't necessarily the best idea. Feels like OEMs are forced to go this way even if objectively as above this isn't actually such a good move.

 

[FusionRedXC60]

Volvo's Tech Day was a good summary of the situation. I've seen this too at the OEMs I work for. I'm not an IT or electrical person (I'm an engine guy) but my understanding is that the status quo is that OEMs buy modules with base software from e.g. Motorolla, Continental, Bosch etc. There's a degree of self development vs complete off the shelf part subtlety between different OEMs. But in general there's an off the shelf module or a partly off the shelf module for all the discrete modules around the car that communicate via CAN or it's successor. The CAN network operating system is e.g. made by the OEM (usually). The bottleneck here is when it takes an age to get to the bottom of if a bug or issue is the OEMs operating system issue or the extra software added to the Tier 1 supplied module by the OEM or an issue with the base software of a module. Everyone does it this way.

Either seeing an opportunity or seeing that Tesla (maybe) took a different approach (just as an aside, Scania and Caterpillar designed and design their own engine ECUs hardware and software in house. Everyone else buys it from Delphi, Bosch or Denso etc) to shift from many discrete modules to a single core computer doing it all. Volvo is going this way. Downside is it's a big task. Upside is you design the hardware and software in house. No Tier 1 base software bottleneck in development or resolving bugs. This route would still have Tier 1s making the modules physically I assume rather than OEMs or Volvo sets up electronic unit production in house.

Geely is for example doing more of this approach via discrete modules with the ECARX division (curiously for me in the UK this company/department has an HQ in London). Geely also moving into making their own chips.

I'm not sure if Tesla doing this (with potential profit margin improvements?) triggered Volvo and others to follow suit or was this parallel independent evolution to take the opportunity to be more self sufficient. Don't know.

Making own chip is not helpful to bring down the cost. CPU/GPU is expensive and only big volume can reduce the cost. It is always easier and cheaper to use existing chip Nvidia if it meets the requirement.

Sent from my moto g power (XT2041DL) using Tapatalk