I'm also aware of a project from a major automotive supplier to attempt the same thing. From my understanding it's unlikely to succeed because manufacturers view suppliers as commodity producers of components they find boring like brakes, steering systems, sensors, transmissions, safety systems, fuel pumps, etc. Not as anything resembling a true partner. Not to mention that it would require competitors to collaborate closely in the production of a highly complex piece of software.
That said, how many times have we seen this story in other industries?
1 Legacy corporation is warned that an integrated, consumer-friendly software architecture for [multi-billion product line] is needed, and failure to produce one creates an opportunity for an insurgent competitor and/or commodification by adjacent supply chain players.
2 Leadership laughs and ignores mounting evidence of just such a threat emerging for up to a decade.
3 Lo and behold, prophesied competitor finally emerges and finds immediate market success.
4 Legacy company announces that they'll bring a competing solution to market, promising investors that they'll produce a similar quality OS, but across 39 models, uniting 457 separate component suppliers AND the entire post-purchase product support infrastructure. They're starting today and promise to launch in 12 months.
5 Legacy company lights billion dollar bonfire to distract investors while CTO frantically tries to source a robust embedded operating system with consumer-grade interfaces and feature set.
6 Best case, no one who has such an OS will license it. Worst case, Google will.
7 Leadership jumps ship, legacy company craters or slowly slides into irrelevance, and CEO later gives interviews about how absolutely no one could have seen this coming, with a sidebar complaining about software engineering salaries.
Honestly, this whole narrative is becoming a bit boring at this point. VW is at stage 5. The fact that its leadership consists entirely of charlatans is self-evident.
No. Volkswagen is at the stage that four years after deciding to bet the company on going electric, deliveries of their new EVs based on their new EV platforms will start next year.
Their new software plan goes hand in hand with their EV push. Volkswagen is going to be the world's biggest producer of EVs because they're spending the most money to be it and they have the scale to execute it:
> The fact that its leadership consists entirely of charlatans is self-evident.
Charlatans maybe but the real problem is that they are old men with ballpoint pens who proudly claim that they have "petrol in their blood".
The real tragedy is that most German automotive managers have no clue about the complexity of modern software and I'd bet most of them would answer the question: "What is more complex, the mechanics or the the software?" with "Mechanics, of course!"
without a blink.
I disagree. A lot of large German auto companies are opening tech hubs here in Lisbon, and from what I’ve heard after speaking to people who work there is that the company is well aware they know absolutely nothing about software and are years behind.
So they said screw it, let’s give engineers/hubs freedom to work with whatever technologies they feel would be best, structure themselves and control their output without micromanagement from above. Basically copy some trendy companies and see what happens.
One of the hubs adopted extreme programming, which I’m not sure about but perhaps has value in their industry. But all in all from what I’ve heard is it’s working well.
There is also a lot of non consumer facing software that compliments a car line. That all needs development and maintenance.
> So they said screw it, let’s give engineers/hubs freedom to work with whatever technologies they feel would be best, structure themselves and control their output without micromanagement from above.
Probably because this is what any sane CTO would tell them. But it probably doesn't help much when the above management tells them their task is to build an operating system. Without first getting a whole lot of technical people's feedback first.
If they pull it off it could be pretty cool, but considering how much they already have to do to adopt AI, I doubt they'll have the talent or know-how to really pull this off.
An interesting question is would it be easier for a tech group to build a better car company, that now suddenly is all about technology and less about designing and running a cost-competitive manufacturing system, than a car company could build a better tech company. But it's clear there's quite a few laws and other barriers to entry to help out the entrenched car companies who have deep local gov ties.
IBM is an example where the quick panic solution (original IBM PC) surprisingly was quite successful for a while but also bit them in the ass in the end (modularity and exclusive use of off the shelf parts).
PCs would have looked much different had Compaq not been so lucky with their reverse engineering attempt, hence that whole failed PS/2 MCA recovery attempt.
The MCA bus fiasco was hilarious. Basically, IBM decided it wanted to retake control of the PC and get everyone to abandon the clones, so they came up with the PS/2 and its proprietary MCA bus, and they really thought everyone would suddenly abandon the clones and their open architecture and buy proprietary IBM PS/2 machines that were incompatible with everything. They didn't seem to understand at all that now the cat was out of the bag, they couldn't put it back in. It'd be interesting to see an interview of the idiot executives that hatched up that doomed scheme.
> manufacturers view suppliers as commodity producers of components they find boring like brakes, steering systems, sensors, transmissions, safety systems, fuel pumps, etc
Does a car manufacturer really find these things boring? Half are like the blood and oxygen of a car, and also includes some territory for unique selling points.
Look under the bonnet of any VW, Audi, Skoda etc, you'll find that all the sensors, valves, computers (and ECU software), filters and a large amount of the parts are made by Bosch.
It'd be interesting to know if they even had the skills in house to build a ICE car from scratch themselves. You can't knock it though, Bosch products tend to be excellent.
The cars are already modular. Internally. There are up to 150 different CPUs (meaning different modules) inside a modern car. All from different manufacturers. See https://en.m.wikipedia.org/wiki/AUTOSAR
>Legacy company announces that they'll bring a competing solution to market, promising investors that they'll produce a similar quality OS, but across 39 models, uniting 457 separate component suppliers AND the entire post-purchase product support infrastructure. They're starting today and promise to launch in 12 months.
Yup. Sometimes the only way to make real progress is to burn it all down and start over.
The big red flag is when companies pre-announce products that will allegedly be released, long before they are real, but well past the due date they should have already been on the market. This whole thing also sounds a lot like VW’s electric vehicle promises.
Why do they do this? Because executives pay is tied to stock price and lots of things can fall apart really fast if the companies cost of capital skyrockets.
It's just my own ignorance speaking (probably), but I can't say that I like the idea of Android getting near/intertwined with mission critical systems. Hopefully 'infotainment' implies something that is well clear of 'stopping' and the like...although I'd just as soon that they dropped the 'tainment' part and simply provided an interface for diagnostics, HVAC, &tc.
What bugs me the most about the 'tainment' part is that not only do I find most of it irritating, but that you are binding technology that obsoletes quickly with an expensive product that should last 10-20 years. It's a shame that car companies have picked this as an area for product differentiation.
Completely agree, especially on the second paragraph - just like how 'smart' TVs ship a long-lasting high quality panel with an obsolete/user-hostile operating system.
On the first paragraph, I thought I'd read this battle is mostly lost already, not with respect to Android, but car software in general is far more intertwined and less robust than you'd think prudent.
It can be quite hard to ignore the smart part. For example, my TV can't even display the input selection menu properly. It moves about as it tries to load an ad, then can't.
Just because you mentioned smart TVs - I had the same opinion, but was then blown away by my new LG OLED TV. It's so freaking fast with Netflix / Prime / YouTube and the UX is really well made (others are probably similar by now). For now I'm really happy with it and I can't see how the software will detoriate the TV so much in the future, that it I will hate the smart stuff.
Yes. It's obviously common practice to partition systems in cars. The infotainment is not running the ABS or airbags.
That said, the people who work on infotainment have to take safety very seriously too. Even just static visual imagery can cause distracting optical effects if it's high contrast or the right shape/pattern. Anything with motion or audio presents very real risks.
If you take the 30k foot view, tbh, the auto world seems kinda nuts. We buy objects that are a sizeable fraction of our annual income if not more, that very rapidly depreciate, and then 20 years later we toss them out as scrap metal to recyclers.
There are many parts to cars. Don't forget the basic ones: exposed to the elements for decades, driven across uneven bumpy bits, even the 'consumable' parts also degrade fairly rapidly...
For an electric, I'm unsure on the long-term life of motors. Batteries are the main consumable, but are actually safer and simpler compared to hot oil and fuel separated by thin gaskets.
Still, even with the consumables exempted, I'm not sure that cars actually seeing road-salts in the winter and scorching sun in the summer last that long (40 years seems optimistic, and I know I HATE being behind any 'classic' ICE that no longer has to pass emissions).
> but that you are binding technology that obsoletes quickly with an expensive product that should last 10-20 years.
Manufacturers want you to have reasons to upgrade. Manufactured obsolescence is absolutely a feature to them. Don't want to keep driving your sufficient vehicle after 5 because "it's insecure and doesn't run the latest OS"? That's a guaranteed revenue stream if they can get enough of the other manufacturers go to along.
The same was probably said 15 years ago about the computerization of the car engine and their rapid deprecation. Did the average lifespan of a car decrease since then? How ridiculous is it to drive a 4WD into the desert, full of electronics that can’t be repaired and where a single chip can cost 300$ and has to be shipped from mainland America...
So yep, with car computerization they reduced the lifespan from 30 to 12 years, and with infotainment they’ll reduce it to 8 years. Same length as a Tesla warranty.
Vehicle manufacturers are just giving customers what they want.
My family recently bought a new car (a Volvo FWIW) and for the first time we really didn't care about horsepower or fuel efficiency or anything really related to vehicle performance. It was all about the electronics and safety features.
There has to be a balance. Electronic fuel injection (EFI) is far more reliable and adaptable than even the best carburetors. Electronic safety systems may have reduced some of the longevity of a vehicle, but the alternative is higher chance of injury or death.
I think that the EFI argument is something of a red herring, no one is suggesting a return to the Quadrajet.
OTOH, EFI can be done as a modular (and replaceable) product rather than as part of monolithic whole-vehicle design, but that last bit of goodness or regulatory need is likely not met.
I just think of it all as being Peak ICE. The last generation of piston engines is going to be crazy complex and probably deserve to be usurped by it's battery-powered successors.
> The last generation of piston engines is going to be crazy complex and probably deserve to be usurped by it's battery-powered successors.
My car-loving boss just got a hybrid RAV4, and he absolutely loves it; combines the best features of both with (presumably) the only downside of lower maintenance (but even then there's less wear-and-tear on the ICE side). I suspect we will see quite a slow, gradual transition through hybrids to battery-only.
Of course, this is Australia (and I'd imagine similar in the US and Canada) where it's more common to drive long distances.
I have a car with EFI and it's 28 years old. We can the best of both worlds. Simple and reliable electronics, instead of a full computer running Android.
I’ve got a 91 Ranger (EFI) kicking around here and have the exact same experience. It has been nothing but reliable in the last 5 years I’ve had it. A couple small mechanical things that had to get changed, but electronically there has been zero complaints.
"where a single chip can cost 300$ and has to be shipped from mainland America..."
that assumes that the failed part is even available.
I'd sure be irritated to have the controller board for a Cadillac XLR convertible top fail and have to shop junk yards, no doubt the wrecking yards are hip to small things of value that they can store on a shelf somewhere.
To be fair, I know more than one purchasing guy, usually in boutique electronic product companies, who spend a fair amount of their time tracking down obsoleted parts on eBay to repair or build one last batch of a metal box with circuit boards inside. It's just the nature of modern times I guess, but luckily we can often pick the era we want to live in in our personal lives.
I just need to stock up on old Thinkpads, avoid home automation, and carefully choose any car purchases.
Unfortunately there are very few car electronic controllers made in the last 20 years where you can replace modules without re-coding them... a $2 part from the scrap yard is useless without the manufactures re-coding software and a couple of hours at the dealership. Unless you're rich I doubt there will be any "classics" from 1990 onwards you can maintain yourself.
This is exactly why, instead of buying a new vehicle, I'm repairing my 15 year-old one myself. I hate the thought of something so expensive and resource-intensive being thought of as disposable.
>some models simply won't run if the infotainment system is broken; the navigation GPS provides the vehicle's master time counter, and without that, the powertrain won't function
Balkanization is about right. In my experience developing driver assistance systems, there are enormous efforts spent on political turf wars instead of proper design and engineering.
This is not surprising, as the team structures, at least as of a few years ago, were traditionally set up with tradtional manufactirng in mind, around parts and control modules, while the functionality exploded (within 1 or 2 generation of cars) and crossed those boundaries without adequate processes in place properly architect the interactions from a bird's-eye view, creating major computing power, network bandwidth, and most importantly to OEMs, cost bottlenecks.
And as cost is king, nobody wants to budge on increasing cost on their own module, and critical architectural decisions aren't made as much as put off until there isn't any other choice left except to hacking in the most critical bits with one eye closed and hoping somebody else's jenga tower falls first.
That, I imagine, is how you get that GPS time thing, I can see how it all started: "Oh, I can save $.30 on my module if I don't put in an RTC, that'll get me a nice bonus for cost saving, GPS is going to have one anyway ..."
I've worked in that environment before. It doesn't help when upper management has no software or hardware background whatsoever (e.g. MBA or "sales" or whatever) and in that case the only way they see things is in terms of dollars.
One time I was working on a new project and I wanted to put an SD card on the board so that we could have a log. I was asked how much the log was "worth" so that they could justify the cost of the extra hardware.
> One time I was working on a new project and I wanted to put an SD card on the board so that we could have a log. I was asked how much the log was "worth" so that they could justify the cost of the extra hardware.
Very much like testing. Incalculable if you need it (and you certainly do), but managers who don't understand, just assume that you're doing something wrong if you need to "waste time" reading logs or writing tests rather than adding features.
I think you described exactly why writing good software is hard. Not because writing good software is an intractable problem, but because politics often takes precedence over it.
In the business world there is a lot of room for BS so people get away with it. But in software, it can doom a project.
This sounds like a swoftware written by an oldschool electrical engineer or something. I mean, in electromechanical systems that do not have computers, engineers find interesting ways to do more advanced things.
For example, ovens with clocks keep track of time through the grid frequency and if the grid frequency changes, the clocks diverge too.
Any time you get cross-disciplinary stuff done by people significantly lacking in experience in one of the disciplines, you get stuff like this. Programmer art (shudder), artist code (double shudder), EE software, software engineer circuit board layout, mechanical engineer wiring... Smart people will figure something out but it always looks janky to specialists in the figured-out field.
And the same applies for people criticizing the fields they know little about. I'm pretty sure that not all electrical engineers in the world are idiots and that there's a solid reason why they've chosen the grid frequency as a standard way of doing it. One thing that comes to mind is that the crystal oscillators frequency is very temperature sensitive, so probably not ideal solution for using in ovens?
Grid frequency actually is hyper accurate in a lot of places. i.e. in the Western US the grid is regulated such that a clock running off it will never deviate more than 2 seconds from the atomic clock derived National Bureau of Standards time [1].
Compare to a typical quartz oscillator which will gain or lose around 15 seconds per month.
This is being pedantic but I guess I was saying it's not hyper accurate in terms of time keeping, even though it's pretty accurate in practical terms.
Also, I guess that means my devices with clocks that drift more than two seconds, ie several full minutes, are actually using crystal oscillators? I guess it makes sense for a microwave which needs second level granularity, but that seems odd to me.
The crystal in your microwave's CPU clock circuit will drift, and so will the time-of-day derived from it until you set the time again, if there is no external reference. The benefit of using the incoming 50/60Hz AC signal as a timebase is that the power station is responsible for controlling the long term frequency to avoid drift so it can be used as a very reliable reference.
But often times the "good reason" is that the engineer can quietly solve a problem in their cubicle without fighting an epic political battle that affects multiple departments. From that engineer's perspective, it's a perfectly rational choice.
And that engineer who writes code might be a brilliant engineer and competent coder, but they just haven't been on many software projects before and they don't see how damning that decision could end up being.
It works and everyone is happy, but it's a wart that people have to build around for the next 20 years.
As a software guy I have to point out that often software "engineers" overbuild and abstract the wrong stuff, doing stupid things like making everything a layer. There is way more bad code written by professional software engineers than by EEs, who tend to be more results oriented.
I partly agree, but also, have you seen EE code (from EEs who are not also software devs)? It's not terrible in a "uses php instead of lambdas", it's terrible in a "absolutely zero consistency in variable naming, indentation, or in fact anything, also any comments will consist entirely of broken code or cryptic half-words."
I deal with this stuff semi-regularly in my day job and wow.
It's because of cost and because grid power is generally very accurate (at least in the US). No one cares if their oven clock loses 10 seconds per year, and "white goods" (appliances) are generally designed to minimize costs as much as possible. The electronics in a typical oven are very simple, and using the power line as a frequency reference is very easy and cheap to do.
Also, crystal oscillators aren't that accurate. You can make them more accurate by heating them to a given temperature, but this is something only done in expensive test & measurement equipment.
There was a really interesting anomaly last year where, due to a political dispute, Serbia and Kosovo refused to synchronize their grids, causing the entire European grid to drop from 50 Hz to 49.996 Hz, in turn causing grid-based clocks to run 6 minutes slow by the time the issue was resolved. [1]
I don't disagree with you that grid frequencies are generally very accurate and they are a practical clock source for cheap clocks, I just found that interesting.
Also they know the incredible effort that goes into keeping the grid synchronized and that if any generator is out by any significant fraction of a waveform then a few fast oven clocks are the least of their problems.
Even just between different SE fields this can happen to a degree. Let a C++ programmer write a website with PHP and Javascript - it will probably work, but I'd guess for a web dev the code would look a quite weird.
and famously if there's ever a blackout, most of modern ovens will refuse to work until you setup the clock... never seizes to piss me off each time I have to do it...
This is a security feature. Most modern ovens allow you to set a timer with auto shut-off. This is very useful if you want to go out and come back to a ready meal. You can imagine what would happen if the power went out for a while and after it came back the oven switched on automatically. Refusing to turn on the oven when the clock is not set is the easiest way to prevent problems when it is in use.
good point... still, quite annoying when you come to check your meal cooking in the oven for the last hour, only to find out that it never actually switched on because of the clock :)
This is purely sloppy. Dealt with a similar issue on a UAV project. We elected to check for a gps time (which fails if, e.g. indoors), and fall back to os system time, with a note in the logs. It can't do anything mission critical without gps anyway, so this only even shows up in dev/debug settings.
I'd guess they use GPS as ntp (rather than direct time source which as you note would dangerously fail in plenty of normal situations).
But since the infotainment already dealt with time, they just used that as a "system clock" straight, so without infotainment there's no "system clock" to be had.
Is your OS time at least battery backed? I can tell you that nightmares start when all your clocks go back to 1970 on boot, and you need a clock to validate certificates on boot...
Not only that, the other providers are extremely deficient compared to Google Maps, from an end-user perspective.
My 4-year-old car came with an infotainment system using HERE maps. It's laughably bad compared to Google Maps, and here's why:
1) The HERE maps aren't auto-updated. You have to go to some trouble to download updates twice a year, and then install them on the car with a USB stick. You only get 3 updates for free, and after that you have to pay a huge price for each update. Google Maps is updated constantly with new roads, construction outages, even when a street is going to be blocked for a parade.
2) The business data on the built-in system is sparse. With Google Maps, pretty much any brick-and-mortar business that exists is on there. And I can easily see the business's operating hours too, and it'll warn me if I'm going to get there too close to closing time.
3) Searching is far easier. Finding a destination on the built-in system is like something from the 90s, and you generally need to start with a state, city, etc. With Google Maps, you just start typing a name and it pops up suggestions, which usually gets you the place you want very quickly.
4) No traffic updates on the built-in system. It has no way to get traffic data; Google Maps has this by default unless you're in a "dead zone".
For anyone interested in mapping technology, check out the book "Never Lost Again: The Google Mapping Revolution That Sparked New Industries and Augmented Our Reality". Written by one of the founders of Keyhole, the CIA-funded (In-Q-Tel) company that created the technology that became Google Earth, the book describes the creation of Google Maps, Google Earth, and the political machinations of different Google departments vying to own "geo".
What competition? There was never anything on that level before Google Maps came around. It's not like Mapquest ever offered similar functionality. Google didn't invent GPS mapping, but they did seem to invent combining it with a bunch of really useful other information like business addresses, hours, reviews, photos, easy searching, etc. And that doesn't even include stuff that isn't useful in cars, like walking/biking directions, public transit directions/hours/etc.
Would we have gotten all this without a near monopoly like Google? The only thing I've seen that comes close is Apple Maps, and that too is backed by a gigantic company with its tentacles in many things. Some company that only does maps isn't going to have access to all that data, so you'll just get a program that makes pretty maps that's only useful if you know the GPS coordinates for something, which of course no one does.
The competition which couldn't improve their products as fast or as well because they did not get free money from their ad business and now had to compete with gratis good quality maps to boot.
For a long time after Google Maps entered the market there were better products available from various providers.
Were they really better though? Maybe by certain metrics, but maybe not by other metrics.
If someone shows me a GPS navigation app that's really pretty, fast, and even shows me where speed traps are, that's nice and all, but what if it doesn't let me just type in a business name, and instead I have to actually know a street address? Then it isn't very useful to me. I'll go back to the app that lets me navigate to business names instead, because that'll save me a lot of time by not having to use a separate app just to look up a street address for every destination I want to go to.
This reminds me a bit about the debate between Google Maps and Waze (yeah, I know they're owned by the same company). The Wazers love waze because it's cartoonish and easy and shows speed traps. That's great, but to me it's too simplistic, it doesn't show alternate routes in real-time (GM will show me a gray route, saying "similar ETA", "3 minutes slower", etc. as I drive), and it's absolutely useless for public transit, so I stick with GM. And this is with two free products both owned by Google.
Finally, what "better products" are you talking about? I remember quite well when GM came out. I switched almost immediately from MapQuest. This was back when it was a PC-only (web) application, of course. At the time, MQ had a clunky interface, and then GM suddenly came out using AJAX, and I could click and drag the map around! It was utterly amazing compared to MQ that you had to use pan and zoom buttons for. Maybe the map dataset was better for MQ? I'm not sure about that, but let's just say for argument's sake that MQ had better map data. That's fine, except that there's more to using a map application than the dataset: the user interface is extremely important too. So this seems like a good example of ignoring how important the UI is, and then wondering why so many people suddenly abandon the "superior" product for the one with the easier-to-use UI, which is something we've seen over and over and over in tech over the decades.
I was using iGO offline maps for navigation on Windows Mobile where one could search for addresses and also POIs. Sygic was a similar app later available on the Sony P1i.
Both of these were offline apps using maps from multiple sources, at a time when Google were still gathering data and mobile online access was expensive and inconvenient. At least when using iGO the iPhone and Android didn't exist yet.
Buying map data is not a problem for a big car manufacturer. They already have their partners for that, they sometimes partially own those partners, and they will port to whatever platform chosen. Google is the weird one in that case, who insist on selling online services, is from what I've heard generally unwilling to commit to long-term setups, ...
The VW guy was talking about the handing over of sensor data to Google in the context of "how much Android" is acceptable, since apparently in his opinion opening up for the MIB is as far as he'd go. The square brackets are Ars Technica's.
The key takeaway for me here is that they want to build an Android-based OS that will run both user applications and, say, your traction control on the same stack. Given the very mixed history of Android security, updates, and more, I simply can't see how this would end well.
The article must be misleading. Android would never have the latency or reliability to run things like traction control and fuel management. The engineers at auto companies are well aware of this, but perhaps the author of the article is not.
Contrary to popular belief most managers in industry are not 100% clueless. No manager in a car company would be so incompetent that they would off-load time critical functions to a phone. My money would be on incompetence of the writer long before I'd suspect the people on the other side of the interview.
I've been through a couple of those myself, it is always very interesting to see how your words come out once they've been interpreted by someone who is essentially clueless but well-meaning and trying to understand something that goes above their normal day-to-day level of complexity. And that's the good case, the one where they don't have an agenda to push.
> Briefly stated, the Gell-Mann Amnesia effect works as follows. You open the newspaper to an article on some subject you know well. In Murray’s case, physics. In mine, show business. You read the article and see the journalist has absolutely no understanding of either the facts or the issues. Often, the article is so wrong it actually presents the story backward-reversing cause and effect. I call these the “wet streets cause rain” stories. Paper’s full of them. In any case, you read with exasperation or amusement the multiple errors in a story-and then turn the page to national or international affairs, and read with renewed interest as if the rest of the newspaper was somehow more accurate about far-off Palestine than it was about the story you just read. You turn the page, and forget what you know.
Well, you do know that the newspaper isn't written by a single person, but by a large team, so I guess the hope is that the journalists who wrote articles about far-off Palestine were more competent than the journalist trying to write an article about physics.
The article does say Android will be used, but it doesn't say it will be used for anything other than infotainment.
It isn't totally clear, but there is one fairly strong hint that there will be separation. They quote Singer (the person in charge) as saying, "What we are now doing with these so-called enabling functions is taking them out of customer functions, putting it in a middleware software layer." In context, it seems to be contrasting that against designs where the car's basic functions were too entwined with infotainment, so "taking out" seems to suggest separation.
The writing in-between the lines is that all car manufacturers want to run everything on a single-board computer. Perhaps the only exception is some things deemed mission critical such as the dashboard - issue being who defines what is mission critical - which will run on a separate "safer" SBC. This allows them to save quite a bit of money per a car, not just in SBCs but also potentially in wiring. Not sure it's a good thing but it is the direction things have already gone.
Cutting out other suppliers will be huge cost savings for Volkswagen if the plan works out too.
As first it's only for the infotainment system. And the rear view camera is already a part it. And then at some point the manufacturer lets you adjust your adaptive cruise control through it. Then more and more critical things are integrated into what used to be "infotainment", until very soon the entire car can't run without the screen.
I think on most cars the infotainment systems runs on a separate computer that is connected to the drive computer via a bus. With a bit of commonsense design it should be fairly simple to have a car thats still useable without the infotainment computer.
I very much suspect that "same stack" here means some blurry "API" and some shared code, not actually the same OS all the way down. Or maybe the same OS at the lowest level, and then running something like Android virtualized for some of the user-facing bits.
I greatly look forward to being able to root any 10yo car and put the traction control in "go forward no matter what" mode rather than "we're cutting throttle so the CVs can outlast the warranty" mode. That said, per the article it sounds like android is just for the infotainment system, not the actual drive-train.
Edit: Since I'm apparently I'm so wrong does anyone want to tell me why?
Being able to use every bit of power the vehicle can put to the ground is going to be harder on parts than just mindlessly cutting throttle when wheel spin is detected. I'm fine with that trade-off if it means I can give my grocery getter traction control that's tuned for maintaining forward progress in low traction (like what most of the new Jeeps do if you set the dial to "rocks") as opposed to preventing wheel spin by cutting throttle (what basically everything else does). CV joints were just an example of a part that might wear out sooner and a hypothetical justification for using the dumb kind of traction control.
I don't know where you learned about traction control, but the normal way it operates is to apply the brakes on the wheel that's spinning so torque is redirected to the other wheel.
That's how it works in theory and did on some of the first generation systems. In practice most manufacturers cut throttle too (assuming throttle by wire, which most new vehicles have) and you often have to pay for some "off road" trim to get the little button that makes it not cut throttle. On a lot of AWD mom-mobiles you can't even get that option. Also, most traction control systems aren't tuned to hit the brake on the spinning wheel aggressively enough (IMO).
The Episode of Dirt Every Day where he takes a Jeep Compass off roading (it's on youtube somewhere) is a great example of what you can do with good traction control.
I get a little bit worried when I read that they want one platform for everything, the security needs of different systems are too different. I'd say a modern car needs at least 3 separate systems:
A low level system for all critical features, this should be coded as safely as possible and run on a slow almost unbreakable computer.
A selfdriving features module, doing all the computations that are too intensive for the low level system. The low level system must be able to detect malfunctions in this system and act accordingly.
An infotainment system, just assume that it has been hacked when designing the other systems. It should be simple to prove that this system cannot take over the rest of the car.
I think this is a result of automotive technical progress slowing down. And I don't mean that in a bad way.
For the past 100 years, we have been developing better and better cars, with both breakthroughs and incremental improvements in the drivetrain, safety, comfort, reliability, etc.
Now though, cars are consolidating. Most car companies only have a handful of 'platforms' on which they build various models. For example: Volvo currently only produces 3 models of combustion engines. We have reached a point where numerous subsystems are just 'good enough' and require little more R&D.
During the past 20 years or so, the software had to keep up with all the new tech coming out in the automotive world. Now that development of automotive hardware is slowing down, it is time to focus on the software for the long run.
Tesla has been doing this for a while now. Their electric drivetrain is at point where there is more for them to gain in software than improvement of the hardware. For other car manufactures it is a bit harder to accept that, as they have been developing hardware technology for a long time.
Edit: fixed my wrong example about Volvo, thanks to C1sc0cat
I wholeheartedly disagree that automotive progress, on the mechanical side, has slowed down. The ride quality, handling, fit and finish, noise & vibe, chassis rigidity, fuel economy, reliability, have all been improving. Go back 10 years, go back 10 years before that, etc., and you will see incremental improvements. Beyond that, look at how much more "stuff" comes in cars today, but for the same amount of money. That means, as we cram more and more tech in a car, the mechanical bits are getting cheaper. The same is true for motorcycles.
What is really going on, is the improvements in the mechanical design are just not as flashy.
There have been significant mechanical improvements under the hood. Dual clutch transmission enabling things like (very quick) paddle shifting has trickled into a lot of cars over the last decade. Gasoline Direct Injection has also seen rapid adoption. Hybrids were a small niche not too long ago, let alone EVs. Turbos seem a lot more common as well.
Dual-clutch transmissions have had a lot of reliability problems, and not many automakers actually use them. CVTs are probably used much more often.
GDI on the other hand has basically taken over; I doubt there's any new gasoline car sold without it now.
Hybrids haven't been doing well, outside of the Prius. EVs are getting more adoption now than hybrids. The problem with hybrids is the cost: you have a gas engine plus a big battery pack and motor, basically two powertrains in one vehicle.
Turbos however have become very, very common as everyone is downsizing engines and using turbos to get more power, while improving fuel economy.
Yeah, I cry every time I have to use some cheaper new rental car compared to my 15 year old bmw 3-series (not even m-series, just regular one with diesel engine on top of that).
Sluggish, puddy garbage of an experience compared to pure joy I can experience with my own.
I would say improvements were done to make things more cost-effective to increase margins, but not necessarily better cars per se.
Other things are with furniture I obtained with an older apartment - amazingly well built, can still hold together well after 50+ years. It was not luxurious back then, but would be definitely be luxurious now.
In a new 3-Series you can turn the steering wheel with your index finger, there is no manual transmission, you hear fake engine noise.
The 15 year old car offers a completely different driving experience and he might prefer it over the bland, comfortable new car.
I'd be curious to see the numbers, but I'm pretty sure the new one has more oomph than the old one. The comfortable aspect I get, but new BMWs have plenty of bite in them.
See those slim runflats? Runflats save space in the trunk for the spare, and them being slim means they are hard. So in order for the ride to be softer, the axle suspension changed, and much softer silentblocks are being used. The result is, that your drive doesn't feel like your car is sitting on rails, like E90 or E46 did, instead it is muddy and jelly.
Another disaster is the start-stop system. When warm, the needles in starter can break. Meaning, that you just went down from highway to the petrol station, refuelled, and you cannot start anymore. You are stuck there on the station.
E90 came with RFT tires and at least E92/E93 lack spare wheel well in the trunk. I dunno about E90, it might be an option there like it was in E60s (but it's very rare).
They didn't use electric steering thankfully. It went downhill starting E9x though, the materials used inside wear much quicker, the suspension and road feel is somehow less than E46 and the chasing of weight savings but increasing comfort through electronics and servos means the car is heavy still but you no longer get that satisfying door or bonnet thump like you did in E39/E38/E46, which were pinnacle of BMWs engineering IMO. Cabin noise insulation is also less than pre 2005, I assume because of weight savings. The post 2005 cars are so complicated that non-enthusiast ownership is most likely a bad experience.
Don't forget the styling. Every car looks like the exact same boring "bar of soap with wheels" today. There's a little variation in terms of grill and light clusters, but from a distance I can no longer tell the difference between any of the top 10 sedan or coupe models.
The new cars have electronic steering, electronic gearboxes, electronic everything. That much tech essentially eliminates "road feel". Some older cars are actually much more interesting to drive, despite their flaws.
Go back to a BMW of 1990, and it's not that much different from a new car today. Fuel injection, air bags, ABS. Internal combustion engine, exhaust, radiator, transmission. etc. etc.
Heck, even back then it was perfectly normal for a car to get better than 10L/100km (23.5mpg)
There have been incremental improvements for sure, but in all honesty it's the same thing, just tweaked a little.
A 1990 BMW 3 Series got 18 miles a gallon and a 2019 BMW 3 series gets 30 combined. 1990 had a 5 speed and the new one has an 8 speed.[1] The power and efficiency of ICEs are night and day vs the 90s.
You could literally same the same thing about anything and it would be a huge generalization.
“It’s just a CPU/RAM/Disk/Nic with a monitor and keyboard. We had those in 1980s and it’s only incremental every since.”
This is like saying if you go back 10 years, computers still had RAM, hard drives, CPUs, and GPUs so they're basically the same as today while neglecting the fact that there have been significant improvements in all of those and a modern computer would run circles around one from 2009.
There aren't many noticeable improvements in cars from one year to the next, but over the span of a full model change, you usually get fairly significant improvements in safety, fuel efficiency per horsepower, and in-cabin features.
Also note that we haven't reached maximum ICE performance yet, either. For example, Mazda is soon to release their Skyactiv-X engine, which is a sort of hybrid spark and compression ignition system with a large fuel efficiency increase over traditional engines.
Try driving one of the latest cars with all modern driver aids etc. for a few hundred miles: lane departure, blind spot warning, parking sensors, automated parking assist, distance sensors etc. Include routing around traffic congestion and a requirement to stop halfway for some unexpected reason, necessitating use of in-car navigation system.
Then try doing the same with a 1990 BMW. Unless you are being deliberately obtuse you will miss the driver aids etc., as they make the whole experience so much easier. Old cars are simply nowhere near the latest cars for safety, comfort, ease of use etc.
Anecdotally I would argue that the last 10 years have seen a more significant improvement in car technology and capability than the prior 10 years.
If we look at power alone even though EPA regulations have only gotten stricter during the same time frame if we look at the same model of car (Ford Mustang GT) we see that it has gone from 310 HP [1] to 460 HP [2] in 10 years. The same power that was delivered by the V8 which powered the Mustangs that my friends and I pined over in High School (ca. 2004) is now delivered by a compact I4 turbocharged engine which also delivers 25 MPG [3].
I think you're mixing up ICE and electric, which is a mistake. You might have a point about ICE, but there's a ton of hardware improvements to come for EVs.
>Most car companies only have a handful of 'platforms' on which they build various models.
Yeah and VW just spent years creating a completely new platform for EVs. Everyone else is going to need to do that in the next decade or so as well.
> [Teslas] electric drivetrain is at point where there is more for them to gain in software than improvement of the hardware.
Tesla are continuously coming up with hardware changes. Models S and X moved to a completely new 'Raven' drivetrain just back in April.
Alongside that, they do try and find software fixes to improve the existing fleet, which I guess is what you're thinking about.
Just to add to this, ICE cars have been though all the experiments about front wheel/ rear wheel drive, front engine, rear engine, air cooled/ water cooled. Those choices are up for reevaluation, plus a few more (how many motors). Its possible that Teslas choice of motor and battery layout will look as outdated as running boards in 30 years time.
Electronic brakeforce distribution means that when the ABS module is not working, the car will brake worse than the same car with no ABS and purely mechanical brakeforce distribution.
This is wrong every time anyone has ever tried to claim it throughout human history for any piece of technology, so much so, that there must be a name for it.
I love the fact that progress never stops. No matter how unimaginable it is that some piece of technology could be improved, it is not only improved upon but it will be dramatically reimagined and reinvented and iterated upon endlessly until it becomes entirely obsolete.
Even when people say “XYZ is approaching the laws of physics for how good it can get” they often turn out wrong, like with solar panels there are several theoretical limits which have been shattered by using multi-layer approaches and mirrored or stacked panels, or ways to excite 2 electrons from one photon, that ultimately push the technology forward in one dimension or another.
How much do you think is spent on perfume / scent science annually?
Besides the core quality and variety of the scents, technology to prevent wick flare-ups, technology to alter burn rates, more economical mixtures of the wax, better composition of the jar to support cheaper shipping and less breakage.
That’s not even to mention all the new competition to the venerable candle; non-burning LED candles, plug-in oil dispensers, time release mechanisms, vaporizers, etc.
I can’t be the only one disappointed in this. Making cars at scale up to modern standards is really hard. Making great software is also really hard. I have serious doubts as to whether a company that does the former can have the culture, management practices, etc. to execute really well on the latter.
The fact that Apple shut down its car division suggests there may have been some issues going the other way.
I wish they’d have moved towards standardizing APIs and “controllers” throughout the automotive world and just encouraged after-market infotainment systems to fill the functionality gap. I just really don’t like that Infotainment ends up having to be a complicating factor in buying a car. It’s hard enough to pick one out that ticks all the functional boxes and feels nice to drive. But then suddenly you get cars that feel great, but neglected the interface design or had a slow UI and suddenly they’ve ruined it.
I think a standard API would be great. "Here's the things the ecu should be willing to expose": basically all the usual EC2 reader registers, as read only.
Here's a standard climate control API, which doesn't touch the ecu stack.
And here's some mixins/ flexible RPCs so vendors can add custom functionality.
Alas, this limits vendor lock in, so that's precisely why I think it's unlikely to happen.
We are talking about the British car industry here, BL is what the the company that made the Mini couldn't work out what profit they where making on the dam thing.
" Chevy Small Block v8 is a little newer starting mid 1950's"
...and not used since 2003. Honestly, it's probably best to view an engine family like this as a form factor, not a design. It isn't like there were any 1955 parts in a 2000 engine, but the interconnects are (mostly) the same. Generally, any 90 degree V8 with a cam in the block is going to be the 'same', designers have pretty much standardized on just a few layouts.
An invariant SBC is like saying disk drives haven't evolved if they use the same mounting points and cabling/logic standards.
I thought they only produced the 2.0L 4 cylinder petrol in various configurations (turbo size, hybrid, non-hybrid). As it turns out they now also have a 3 cylinder engine.
On the other hand, look at an actual car from the 1950s.
Yes they have 4 wheels, wind screen, engine, but reliability is massively different, power is massively different, comfort, crash safety, price (for what you get).
ICE engines are not innovating, it's true, but the context is not explained in your post. ICE engines are going to be replaced by electric motors, which are far simpler to maintain and source parts for, and are inherently (due to physics) much more energy-efficient. Given climate and potential revenues for new ICE R&D drying up, you're going to see all the research dollars in 1) batteries and electric drive-train, and 2) software like autonomy and infotainment and apps.
if you call 1919 a recent innovation - VVT first appeared in ICE then, and made a resurgence in 1958 and again in the 1990's. electric superchargers have been around for a while as well, though not as prevalent due to the energy requirements. changes in the car electrical systems themselves have been allowing those to move forward, not innovation per se.
Both have been overtaken by events. They're not going to show up in many mainstream cars before ICE becomes obsolete. But TVVT might be very useful in applications where electric motors won't yet work. ICE airplanes, for example.
My two cents:
- I'm asking myself why there's no reference to Autosar (https://en.wikipedia.org/wiki/AUTOSAR) which is not an operating system (whatever this means in this context), but yelds a sort of a standard architecture in the car and among third parties. It's a real mess, but it works in this sense.
- rumors exist about VW and other manufacturers pushing for using ADA (more specifically Spark), anyone aware of this?
Yes, it is a mess. At over 15.000 pages, AUTOSAR specifies not general utilities, libraries, battle tested data structures used in the automotive field or whatnot, but actual automotive components that in theory can be developed independently and combined to build the whole car. XML being the medium of choice for this standardization doesn't help either. You're left with tools upon tools that modify XML files and generate C code based on those files and very little opportunity to actually look/modify those C files by hand, even in trivial circumstances. In short, the AUTOSAR idea, from what I've experienced, is to hire an army of mouse-clickers that can use shiny tools to assemble and configure every aspect of the car. They would not learn C or any programming, but learn the actual standard and know what you need to click or check in each and every instance. The tools would then take care of bringing in the code and generate the header configuration files for you.
Anyway, not mentioning AUTOSAR might be a sign that its days are numbered, perhaps?
Then, from my experience, when the mouse-clickers are done, you employ an army of C-coders to work around and/or abuse the (expensive) AUTOSAR components until the system (somewhat) does what you wanted it to.
In my case it was AUTOSAR in name only. Our higher bosses thought we were running AUTOSAR, but the lower you went, the understanding was firmer and firmer that what we actually ran was custom software.
Oh, the days numbered part - nah, I don't think so. AUTOSAR is for very low level stuff, I think unholy combinations will live on for a long time. AFAIK all the component (like hardware, Bosch for instance) vendors only provide AUTOSAR components for integration. Not sure, it was a few years ago I was in the loop.
> Anyway, not mentioning AUTOSAR might be a sign that its days are numbered, perhaps?
I'd like but it doesn't seem so. We're having pending requests for working on AUTOSAR-based software, and a few are for brand new car models. Different opinion regarding EV: here I'm not aware of any single AUTOSAR based element but I may be wrong.
I immediately searched about this connection, and saw a PDF announcement from 2017 where AutoSAR announced that VW was going to start using their latest, greatest version. I can only assume now that either VW found that it just wasn't up to the task of future proofing the platform, or they're trying to negotiate better deals in the AutoSAR space. (I know very little about this market.)
What's truly scary to me is that VW says they'll need 5-10 THOUSAND people working on this. Given that they're starting with Linux/Android, and have all the mature tooling in that space, why should it take that many people to adapt it to a car? That seems excessive.
I wonder if this is fallout from Tesla. They must have a huge advantage in a way here having a clean sheet architecture only a decade old. Between this and superbottle I think the idea to use suppliers as little as possible may pay off.
Over the air updates is vastly under estimated by many people. It is just as much of a revolution in the automobile industry as the electric drive train. As more cars move to electric and electrified drive trains the value of over the air updates that Tesla employs will become more evident.
it frees you from being forced to buy a new car to get new and or improved features. plus building a system like that allows Tesla to add new hardware to their cars with ease as the mindset is in place already.
A model S bought five years ago enjoys most if not all the updates a newly purchased model S has today. that is not something any other maker can claim and I doubt that any even want to go there because not only does it free the consumer from having purchase a whole new car for exciting features but obligates the maker to providing updates to an existing car.
An Android-based infotainment system developed in collaboration with Google has already been announced and demonstrated in 2020 Polestar 2 EV, Volvo's alternative to Tesla Model 3, which might have inspired this "Volkswagen's bold plan"[1]. I have nothing against it, if the update policy would be similar to Android One's. Otherwise, it can easily become a security nightmare. Imagine, if your 5-year-old Audi car was as secure as your 5-year-old Samsung Galaxy smartphone.
Iteresting. On the user-facing side I hope the choice for Android will mean GPL compliance and a way to sideload apps. Can't wait for a LineageOS variant for cars.
My experience with user-facing Volkswagen software has been terrible so far, so I hope a new mandate might improve that area drastically.
As far as the underlying new operating system Volkswagen is developing for all their cars, seems like a smart move as long as the design by committee syndrome is somewhat curtailed.
Despite I love to hack this kind of things, I really hope automakers will enable verified boot / secure boot / trustzone (or whatever the silicon provides). Having someone to remotely install malware on an ECU is scary.
I fail to see how an operating system causes 2 or 3 decades of divergent software engineering to coalesce. That's only probable if numerous components descend from the operating system group in an architectural style that is unified, congruent, and global. That's way beyond the bounds of what we traditionally call operating systems. You are more likely to wind up with a cargo cult.
A big driver not mentioned is differentiation in the market, now that the trend is for platform sharing across manufacturers. Essentially the same car being sold by several different manufacturers with the only differences being the badge on the back and things like the infotainment system fitted. By carving out a big ecosystem early they’re getting a head start in that area.
We're changing to auto market into the cell phone market? Oh no.
"Well, I get Free Miles(tm) with the Chevrolet, but it throttles at 70MPH and the radio only plays Pandoa. Or I can get the exact same car from Crysler, with no throttling, but the miles come in bundles of 1000 and radio apps are $5/month."
The article mentions they will use Android as the base platform.
> Senger also revealed that VW Group will be using Android for future versions of the MIB infotainment platform, in large part because of the robust third-party app ecosystem with that OS versus Linux.
Huh? Android is running on a modified linux kernel. But I get what he means...
There's a competing automotive tech stack called "Automotive Grade Linux" which is being adopted by most of the Korean automakers, many of the Japanese ones, Daimler, etc. That implementation goes far beyond infotainment and the head unit, though it has hooks for both so they might not be in strict competition. VW potentially can just use their Android-based infotainment system on top of a different stack.
I can see what VW is trying to do here, and honestly out of the various rental cars I've driven over the past 6 months, my Golf R's Android Auto/Car Play integration has probably been one of the best. VW also has a history of people doing minor modifications of their cars electronics, both via the intended menus and for more complicated things, via the outstandingly named vagcom (VCDS)*
I think this is the direction more automakers should go. I drove a friends Tesla the other day, and it shares some the same features, allowing you to make fairly specific customizations of certain things through the built in menu.
*Some items which can be customized via the VCDS - DRL's, Windows Auto-Up/Auto-Down/etc, and a bunch of other things I can't remember anymore.
With all the car operating systems (OS) that will no doubt be connected to the internet AND abandoned in the coming years. We need a car OS that is fully open source so companies can offer security updates after they are abandoned, so they don't get hacked and taken over as a full autonomous bot army of cars.
Once an OEM opens up its vehicle data and control to Android (and/or Google Automotive), it's game over for them. The OEM becomes a commodity hardware maker. Plenty of analogy with mobile phones manufacturers. The OEM will not make a dime after the sale of the vehicle while Google will build and strengthen a development platform to enable third-party applications and services and profit from them (something Google knows how to do really well). Think Turo/Getaround, cleaning, refueling, charging, insurance, package delivery, in-car applications, etc.
One way out of this is for the major OEMs to band together and create their own standardized platform that works across OEMs. At minimum, that platform should expose a single standard interface to all third-party service providers.
>One way out of this is for the major OEMs to band together and create their own standardized platform that works across OEMs. At minimum, that platform should expose a single standard interface to all third-party service providers.
This already exists, and it's why there's a slew of Bluetooth-enabled apps on the app stores that let you self-diagnose your own car. It's called OBD-II and it's federally mandated on every vehicle since 1996 in the US and since 2003 in the EU: https://en.wikipedia.org/wiki/On-board_diagnostics
OBD-II isn't actually a real standard. The OEMs all tack on additional stuff to it, so none of the implementations are fully compatible with each other. Ford and Mazda, for instance, use a medium-speed CAN bus that other OEMs don't, so most OBD-II readers can't access anything on that bus. Also, the whole point of OBD was supposed to be standardizing diagnostic messages so anyone could read the trouble codes. However, all the OEMs have lots of proprietary codes that aren't in the standard. The whole thing is a mess.
Finally, OBD-II isn't a standard bus anyway. It's an interface to the end user or technician. The various modules on the car are interconnected with a variety of buses: CAN (high-speed or low-speed or medium-speed), LIN, MOST, etc.
The OBD interface is too cumbersome (vehicle owners have to mess with hardware installation, etc...) and is more geared towards aftermarket devices and software. This effort must be at the core of each OEM and be completely seamless to the end-users and even third-party developers.
In some ways I wish they abandoned infotainment systems completely and just relied iOS/Android to provide those services. Stick a dumb screen in the car with USB/Bluetooth connectivity to my phone.
We were already going in that direction, but perhaps EVs will push us over the edge. Tesla has certainly made it damn near impossible to work on their cars.
Between the need for expensive training and diagnostic equipment, and the ever-widening view of things that are .gov mandated and covered under smog or safety laws including spare parts sourcing, we are largely there.
I'll counter. In a world of electric cars (which I'm in favor of), how long until the politically well-connected emissions testing industry becomes repurposed into mandatory safety inspections at the same or higher price?
My cynical side tells me that if Germany weren't one of the world's largest car exporters, the EU would already be legislating to harmonise these systems globally.
VW is looking for a strong 3rd app support; I doubt pure OS is in the same ballpark as Android for ready-to-install applications from large developers like spotify, tomtom, etc etc
You will have some ISO26262 compliance managers, tons of safety managers, many many documentation writers. This software also must be tested with corresponding hardware, couple hundred people will look after this hardware alone. It’s not another app, it’s software tied to hardware and this is always slow and expensive.
The Apollo Guidance Computer had 36,864 16 bit words of rom [1]. I don't think there's really enough there for 10000 people to get their teeth into. Although it was core rope memory, and it's unclear to me whether you would include the weavers in that total.
This sounds like an underestimation to me. An ECU in car can consume up a team of 200 to 250 engineering resource. This only for those working directly on the ECUs. Add to it the secondary functions, the numbers will explode.
I worked on the infotainment part of recently released top brand car's software. The number of engineers on that was flabbergasting. Way higher than your number.
But the reason was not that it was needed.
The reasons were:
- The approach to developing software (pure waterfall, with lots of agile BS bingo terms as seasoning). As someone else mentioned: Old car companies do not understand software. I may add: At all.
- Upper management throwing more resources at missed deadlines (that were moving all the time anyway). Every seasoned developer knows that more developers will slow you down. Nine women can't have a baby in one month.
- Trying to understand the issues with the project getting pear shaped by looking at burn down charts.
Once I was on a way to one of these meetings in an elevator and someone said: "Gentlemen, are you also on going to our weekly 'Men who stare at graphs' ritual?"
Most senior engineering folks agreed that the work hundreds of developers, dozens of engineering managers, PMs, POs, agile coaches and god-knows-what-other-fancy-title people were doing could be done by a team of around 25.
I completely agree and as such parent is right. 5k-10k people is complete overkill. But then -- given how these companies work -- it is not. It is indeed an underestimation and the very reason this will go nowhere.
A few cars will be released with this OS and then it will be replaced by something much better that someone else, not an old car company, was doing in the meantime.
I have also worked in the infotainment space, now in powertrain space. It seems that the infotainment teams seems to be 2x to 3x the size of a typical automotive engineering team.
Sometimes, I feel that infotainment products are more subjective than the other car functions. eg, Does that "ding" sound notification reflect the values of our brand vs Can the maximum power be transferred to the rear wheels in the snow within 300ms, subject questions vs objective questions. Subjective questions needs more study and analysis, leading to more engineers.
I worked on firmware at Tesla. My estimate is absolute maximum 10 HW engineers, 10 SW, 10 test, a handful of managers, requirements engineers, etc. That would be for an important ECU.
I developed the FW for a couple ECUs singlehandedly along with 1 HW engineer and an intern.
Maybe that's the difference between Tesla and the others.
Even when you look at bigger teams there are still only a handful of people doing actual work. Out of that team of 10,000 I bet 90% of the commits will come from 100.
Not limited to car companies btw. Make that pretty much any industry where software isn't their core business but just a checkbox to tick on their product.
I don't know, how many cars do they have on the road? Must be at least 50 million or so. So 10,000 coders for 50 million users. And these users are, you know, driving cars around, so its kindof important to get the software right.
This sounds like a smart move by VW. Essentially Tesla are showing the way forwards here, being the first automaker to take software seriously.
The number of engineers working at Facebook is in the same ballpark and they have a few billion users. At the same time the number of car platforms is increasingly consolidated and in complete control of the manufacturers.
> And these users are, you know, driving cars around, so its kindof important to get the software right.
That's not something you achieve by throwing more engineers at the problem.
The dash for a car has over 20 software engineers working on it if you count QA. That's a simple component...
Now multiply that by every component that has a CPU in it (every electronic component in the car) and getting to 10k software engineers as seems quite easy.
That said, how many times have we seen this story in other industries?
1 Legacy corporation is warned that an integrated, consumer-friendly software architecture for [multi-billion product line] is needed, and failure to produce one creates an opportunity for an insurgent competitor and/or commodification by adjacent supply chain players.
2 Leadership laughs and ignores mounting evidence of just such a threat emerging for up to a decade.
3 Lo and behold, prophesied competitor finally emerges and finds immediate market success.
4 Legacy company announces that they'll bring a competing solution to market, promising investors that they'll produce a similar quality OS, but across 39 models, uniting 457 separate component suppliers AND the entire post-purchase product support infrastructure. They're starting today and promise to launch in 12 months.
5 Legacy company lights billion dollar bonfire to distract investors while CTO frantically tries to source a robust embedded operating system with consumer-grade interfaces and feature set.
6 Best case, no one who has such an OS will license it. Worst case, Google will.
7 Leadership jumps ship, legacy company craters or slowly slides into irrelevance, and CEO later gives interviews about how absolutely no one could have seen this coming, with a sidebar complaining about software engineering salaries.
Honestly, this whole narrative is becoming a bit boring at this point. VW is at stage 5. The fact that its leadership consists entirely of charlatans is self-evident.