Tesla teardown finds electronics 6 years ahead of Toyota and VW

[jeremy]

Yes, I agree with your assessment of the article. I guess my confusion is better stated as: if the German way (or Japanese way) produces a more reliable system, then why shouldn’t we stick with that? I think the author was hoping that people would just assume that having more people involved is a less efficient way to produce an ECU, but unless you actually compare the new ECUs against the existing ones, how can we actually know this is the case?

[BU508A]

I think, since these ECUs are very complex systems, there is only one way to find out: it must be done and see what happens. Not sure, if I'm very happy with this way.

[nctnico]

I have no experience with Tesla cars, but what is wrong with old designs? Just because a design is newer, doesn’t make it better. Look at the 3458A for example; if it ain’t broke don’t fix it.

I would have liked to see actual criticism of specifically what makes the Tesla system better, not just “Tesla future, Toyota old and smelly”. I still have no idea which is better after reading that article 

My understanding of the article is not, which is better, Tesla or Toyota / VW.
My understanding is: what makes this to the supply chain for the companies which produces automobiles and their suppliers. In Germany, lots of jobs are in the supplier sector for the automobile branch. If those suppliers aren't needed any more, this would have a huge impact to the economy, not alone in Germany.

I think you have to look at the supply chain differently. In the end the supply chain has to deliver what the car manufacturers require. You also have to ask yourself: how much development needs a car manufacturer do themselves and what can they outsource? If they outsource they can spread the development costs of common components which don't add any value to the brand (like ABS pumps, windshield motors, radiators, etc) over much more cars than the manufacturer makes themselve. And it is not like the suppliers are sitting on their asses. They are actively developing new technologies too.

[Siwastaja]

Comma.ai is not a redundant setup.

Yes, sure. But it shows just how much is possible using a single processor and a simple camera. Putting 5 of those processors in a car is cheaper and probably better than designing a crazy AI ASIC.

I'm 100.00% sure whatever "ai" self-driving system with a single camera and mobile phone CPU will fail miserably to provide anything else than technological demonstrations.

There is a long, long way from a demo or even test drives, to a reliable and safe self-driving car. The difference between good conditions and bad conditions is easily 3-4 orders of magnitude in complexity, it's a massive showstopper for most startups on this field.

In more demanding conditions, even human brain with 90 billion neurons and 1000 megapixel stereo vision sometimes have severe difficulties processing where the road actually is. It needs a combination of very sophisticated sensing and very complex processing, including "AI-like" parts like neural networks, but also static algorithms.

I strongly suspect even Tesla, with their custom processing core and a dozen of cameras, will still be a long way from reaching reliable fully self-driving system. They are far from having "too much" processing power, likely still the opposite! I think going custom ASIC for processing is likely exactly the right direction. There will be practical limits in power consumption and parallelization, no one's going to put 5kW worth of server PCs or graphics cards in their cars then spend time synchronizing parallel processing. If Tesla can achieve the same processing power in a few hundred watts, they are years ahead on the processing hardware side. This "version 1" likely isn't what actually brings us fully autonomous cars, it could be version 2 or version 3; maybe it's in 2030's; but if no one else is solving the processing problem, then Tesla is likely the one who gets the fully working system first.

This said, processing hardware isn't everything. I also strongly suspect Tesla, at some point, will admit they need to throw some actual distance-measurement hardware, lidar, even radar, at the problem, even if they currently seem to give the impression they are only working with standard 2D multicameras. Stereo cameras can go a long way, but augmenting it with actual distance measurements will provide a lot of confidence in difficult corner case conditions, and for Tesla, developing a low-cost solid-state LIDAR in co-operation with some innovative player in photonics field is not unlikely at all, even though if they don't publicly talk about it now.

For example, a stereo camera vision will fail miserably to provide any sense of 3D shape with fresh, smooth snow, in uniformly lit cloudy day. I have actually witnessed a human drive directly in a 1.5 meter deep ditch, nearly at walking speed, simply because everything in our 2D vision is pure smooth white with no texture; human brain has hard time figuring out how to stereo image it, I'm sure Tesla will have hard time as well. OTOH, in similar conditions, a 3DTOF camera worth of $20 components does great job producing an accurate (within 10cm) point cloud from the whole site.

Of course, even a LIDAR wouldn't see what's inside/below the snow... Which is again why processing/AI is so important. And you can't do it with a few hundred neurons. An ant cannot drive a car, nor can your dog. We need some human-level intelligence here, in worst cases.

[coppice]

That's a very incompetent article.

Its what passes for journalism in 2020.

The traditional Carmakers usually never design electronics by themselves (any more), they define the architecture in the car, and order the electronics from big suppliers, like Denso, Bosch, Valeo, Continental, and so on.

That's not really true. Some of the biggest car makers are more integrated than appears at first sight. Does Toyota make its own gearboxes? Kinda, because they are the majority shareholder in Aisin. Does Toyota make its own electronics? Kinda, because they own a big chunk of Denso.

Their processing units (µP) in the different modules are of course automotive specific designs to fulfill AEC Q100, for example, and are tailored to the required features, performance and price. Very rarely commercial off-the-shelf processors are used.

Most car makers partition their electronics so the most difficult things to get AEC Q100 qualification for are outside the safety related areas, and don't need it. AEC Q100 is a difficult market to be in. It demands a lot from the vendor, yet prices are very low. Excessive integration works against the car makers.

[SilverSolder]

I have no experience with Tesla cars, but what is wrong with old designs? Just because a design is newer, doesn’t make it better. Look at the 3458A for example; if it ain’t broke don’t fix it.

I would have liked to see actual criticism of specifically what makes the Tesla system better, not just “Tesla future, Toyota old and smelly”. I still have no idea which is better after reading that article 

My understanding of the article is not, which is better, Tesla or Toyota / VW.
My understanding is: what makes this to the supply chain for the companies which produces automobiles and their suppliers. In Germany, lots of jobs are in the supplier sector for the automobile branch. If those suppliers aren't needed any more, this would have a huge impact to the economy, not alone in Germany.

The work that needs to be done won't disappear, though.  The workers would simply be working for another company.

For commodity products, the shared supplier model can be the most efficient.   For example, Ford and GM share a transmission (gearbox) design these days...   it is cheaper for both companies and they are not competing on gearboxes.

[TK]

The main difference between Tesla and the comma.ai solution is the ability to drive in NOT IDEAL CONDITIONS.  Poor visibility and any type of obstruction can affect the performance of OpenPilot.  With just a front camera and maybe the backup camera, it cannot make line changes unless the driver confirms it is perfectly safe.  OpenPilot is nothing more than a smart cruise control and it will never be able to self-drive.  So $999 for a bunch of wires and the Panda device (STM32 chip) is a very expensive solution for a cruise control.

[thinkfat]

The work that needs to be done won't disappear, though.  The workers would simply be working for another company.

Not the same workers, I'm afraid. Maybe not even workers in the same country, or continent.
German automobile industry definitely needs to keep pace with global technology trends in order to be profitable.

[tszaboo]

Tesla has Level 2 self driving, and with current hardware they probably could get level 3.
Audi A8 has certified level 3 self driving, and they are working on level 4. Level 4 has geofencing, most of the time it can do everything and you can take over if you feel like.
The hardware costs are like a used VW Golf in the Tesla. I guess most car makers are not going to invest this much on experimental tech on production cars, because customers want something affordable. It is not the question of "can I do this", more like "does it make sense to do?"

[SilverSolder]

[...] It is not the question of "can I do this", more like "does it make sense to do?"

A lot of research into advanced technologies may only become financially viable many years from now...   

[wraper]

Audi A8 has certified level 3 self driving, and they are working on level 4. Level 4 has geofencing, most of the time it can do everything and you can take over if you feel like.

And it works in their yard, maybe. Since they never bothered to release it to the public. And even a few videos where it is demonstrated by their employees show it's a joke.

[peter-h]

I agree totally with Siwastaja above.

In a presentation on electric vehicles I went to recently, the guy (who is big in this field) said that nobody in the industry has any idea how to make a self driving vehicle which can do what human drivers can do in good conditions.

Even if they had 1000x the CPU power they would not know how to do it.

And then you get less than favourable conditions, bad light, snow, etc. Any skiers here? Try skiing in the shadow of a hill, if the snow is smooth. It takes a lot of cues. Even parking in a supermarket car park is orders of magnitude beyond today's software expertise.

[thinkfat]

Audi A8 has certified level 3 self driving, and they are working on level 4. Level 4 has geofencing, most of the time it can do everything and you can take over if you feel like.

And it works in their yard, maybe. Since they never bothered to release it to the public. And even a few videos where it is demonstrated by their employees show it's a joke.

Heh, Teslas level 2 self driving demonstrably doesn't work either and it's on public roads. 

[schmitt trigger]

We may argue here until hell freezes over, but the fact is that AI, specifically AI applied to the automotive sector, will be the next technology differentiator which separates the prosperous from the struggling countries.

With the German auto sector -for the time being- still considered the world's premiere, and such an important part of the German economy, it would be darn foolish not to invest in AI research focused to the automotive sector.

There was even an article on the Deutsche Welle discussing this topic. It focuses on AI as a whole, not only automotive AI.
https://www.dw.com/en/skeptical-germany-lags-behind-on-artificial-intelligence/a-51828604

Whoever develops an automotive grade AI "standard", will lead the world's industry for the next decades.
As an example of automotive technological leadership, just look at CANBUS.

[wraper]

Audi A8 has certified level 3 self driving, and they are working on level 4. Level 4 has geofencing, most of the time it can do everything and you can take over if you feel like.

And it works in their yard, maybe. Since they never bothered to release it to the public. And even a few videos where it is demonstrated by their employees show it's a joke.

Heh, Teslas level 2 self driving demonstrably doesn't work either and it's on public roads. 

[nctnico]

We may argue here until hell freezes over, but the fact is that AI, specifically AI applied to the automotive sector, will be the next technology differentiator which separates the prosperous from the struggling countries.

With the German auto sector -for the time being- still considered the world's premiere, and such an important part of the German economy, it would be darn foolish not to invest in AI research focused to the automotive sector.

But what problem should AI solve? It reminds me of an example someone told me years ago. Researchers spend decades on developing a robotic hand which can lift an egg without crushing it. Meanwhile in the egg-packaging plants they use robots with vacuum suction cups to lift the eggs and put them in boxes at an incredible speed. Most self driving systems seems to be aimed at using visual information to move around in traffic. There are much better approaches to make self driving cars. The more general direction the car industry is going is to make vehicles communicate between each other and with the infrastructure. If cars can simply tell each other where they are and traffic lights tell cars it is red or green (and this technology is already being deployed!) the need for processing visual information through AI is much less necessary. In the end all a car would need is to be able to detect non-communicating objects. But not to a high degree because information about such objects is also shared between the cars.

[wraper]

If cars can simply tell each other where they are and traffic lights tell cars it is red or green (and this technology is already being deployed!) the need for processing visual information through AI is much less necessary. In the end all a car would need is to be able to detect non-communicating objects. But not to a high degree because information about such objects is also shared between the cars.

And you just need to replace all infrastructure and keep 100% of pedestrians and usual cars away from the roads. Not to say have 100% accurate mapping everywhere.

[Marco]

Has it "learned" not to crash into firetrucks yet?

[BU508A]

But what problem should AI solve? It reminds me of an example someone told me years ago. Researchers spend decades on developing a robotic hand which can lift an egg without crushing it. Meanwhile in the egg-packaging plants they use robots with vacuum suction cups to lift the eggs and put them in boxes at an incredible speed. Most self driving systems seems to be aimed at using visual information to move around in traffic. There are much better approaches to make self driving cars. The more general direction the car industry is going is to make vehicles communicate between each other and with the infrastructure. If cars can simply tell each other where they are and traffic lights tell cars it is red or green (and this technology is already being deployed!) the need for processing visual information through AI is much less necessary. In the end all a car would need is to be able to detect non-communicating objects. But not to a high degree because information about such objects is also shared between the cars.

I think, the AI is aimed to solve the last 0.1% situations, which weren't thought of. The unexpected things which are going to happen and
which can be easily solved by a human with all his knowledge about the world and the rulesets he is following in a flexible way and his experience.

This is a very simple example imho:

[nctnico]

If cars can simply tell each other where they are and traffic lights tell cars it is red or green (and this technology is already being deployed!) the need for processing visual information through AI is much less necessary. In the end all a car would need is to be able to detect non-communicating objects. But not to a high degree because information about such objects is also shared between the cars.

And you just need to replace all infrastructure and keep 100% of pedestrians away from the roads. Not to say have 100% accurate mapping everywhere.

You'd be surprised how well governments keep track on sizes and location of roads. How else can they be managed? Currently GPS is useless for self driving cars but there are terrestrial based location systems at the horizon. For sure you can always find an uncharted dirt road but on such roads there is very little added value of having a self driving car anyway.

[Marco]

We may argue here until hell freezes over, but the fact is that AI, specifically AI applied to the automotive sector, will be the next technology differentiator which separates the prosperous from the struggling countries.

It sells Teslas for the moment and other companies will try to follow their example, but the emperor has no clothes. At some point there will be some government employee in between the next stopped car and the Tesla crashing into it and then it will all come tumbling down.

If you let the auto drive without human intervention for 99.9% of the time, the human won't be paying attention for the 0.1% ... it's as simple as that.

[coppice]

Audi A8 has certified level 3 self driving, and they are working on level 4. Level 4 has geofencing, most of the time it can do everything and you can take over if you feel like.

And it works in their yard, maybe. Since they never bothered to release it to the public. And even a few videos where it is demonstrated by their employees show it's a joke.

Heh, Teslas level 2 self driving demonstrably doesn't work either and it's on public roads. 

That guy is seriously delusional. He keeps telling the car to change lanes and change speeds, including a key lane change to get it prepared to simplify an automated exit from the highway, and then tells us he's not giving the car any input at all. The car does many stupid things during that journey, especially the multiple unprovoked lane changes in the latter part. If you've never seen a car do any automated driving its cool to watch, but its so far away from safe automated driving that Tesla's claims to have automation feature complete by now (which is what they claimed a year ago) are a joke.

[wraper]

especially the multiple unprovoked lane changes in the latter part.

I guess you missed that he set lane change to "Mad Max" mode.

[thinkfat]

especially the multiple unprovoked lane changes in the latter part.

I guess you missed that he set lane change to "Mad Max" mode.

Yes, but the car changed lanes with no reason, with no cars around and nothing slowing it down that would prompt it to overtake...

[SilverSolder]

especially the multiple unprovoked lane changes in the latter part.

I guess you missed that he set lane change to "Mad Max" mode.

Yes, but the car changed lanes with no reason, with no cars around and nothing slowing it down that would prompt it to overtake...

That's exactly what Mad Max would do...