German regulator says it discovered new illegal software on Daimler diesels
wsj.com
wsj.com
There are so many valuable lessons to takeaway from this, but weirdly for me, it's to never depend on innovation. Especially specific innovation.
It's important to remember that everyone a decade ago was riding high about the potential for diesel. Consumers, car producers, and regulators. And everyone was making promises about how good diesel was going to get - based on the expectation that you can just dump resources into a certain technology and get innovation, magically and predictably. But it feels like everyone just hit the innate limitation of the resource and tried to pretend it didn't exist.
Anyway, remember VW themselves had engines that met the spec just fine. They only cheated so they could omit the AdBlue system.
They specifically set goals and policies to make diesels a larger presence (hence why so many power diesel sedans and small cars were made for the EU market but not the US).
The worst you can really accuse them of is lax oversight.
It would be so much better if we just started taxing externalities like carbon or fertilizer runoff and let markets decide the tech to do it organically.
But let's not let the car oligopoly completely off the hook. At some point they all took that step over the line.
1) even if powered 100% by coal plants, these plants have an higher efficiency than cars' internal combustion engines, especially if you consider the energy/co2 emissions due to the distillation process of petrol/diesel. Moreover, having a huge central plant means that you can install heavy and big filters and possibly re-capture some of that co2. You can't do this with cars
2) not all the power comes from coal plants, some from nuclear, some from natural gas, and even from renewables sources
3) when we finally crack nuclear fusion / thorium reactors / we found a way to store the surplus of renewables for underperforming times in a good way (like with molten salt https://newatlas.com/mit-molten-salt-battery-membrane/53085/ ) we can be carbon free from as soon as these technologies spread, without having to way for consumers to change cars
4) Pushing people to use newer technologies means that companies will spend more in R&D on this subject. Imagine if someone finally invents a battery with energy capacity of 5x current Li-ion. A car like the model 3 will have ~ 1500 kms of range
...
Of course there are points in favor of petrol/diesel cars, like the fact that producing current batteries is not that good for the environment, but if we don't invest on this we'll never have the incentive to overcome these problems.
> "Imagine if someone finally invents a battery with energy capacity of 5x ".
3 & 4 are exactly examples of the original point I was making. We shouldn't plan out decisions today based on the lottery machine of future discoveries. None of these advancements are guaranteed.
In the 1910s, steam power was the most promising tech for the development of the automobile. It's hard to imagine a world where we could have just dumped enough R&D into steam power to overcome the natural shortcomings.
more likely it'll have ~ 700km of range, but only cost half the price!
Yes they have incredible thermal efficiency, but two things come to mind: The grid does not, and coal has particular difficulty in controlling particulate emissions. This is particularly true of brown coal, which has a very high water content. Just to be particularly frustrating, coal is not distilled the way diesel is (which is famous for particulate emissions) so it tends to have other stuff like excessive sulphur. Though in that case, that stuff ends up in the bunker oil that goes in big ships.
It wasn't that the regulators were 'picking winners from specific technologies' as you claim, but they were favouring any vehicle with a low CO2 emission per kilometer. If that was achievable by a petrol car, then it was rewarded; low emission petrol cars got the same break, but they're inherently less efficient due to the lower compression ratio of a petrol engine.
Diesel comes with a higher energy density which means there are more CO2 emissions per burned liter of Diesel in comparison to gas.
Burning a liter of Diesel emits 2.6kg of CO2, gas 'only' emits 2.3kg of CO2 [1].
Plus, a Diesel engine is heavier compared to a gas engine with comparable specs.
[1] https://www1.wdr.de/wissen/technik/kohlenstoffdioxid-sprit-1...
However, what is certain is that the CO2 savings from diesel is small and certainly less than 50%. That's not going to get us where we need to go in terms of carbon emissions.
> but they were favouring any vehicle with a low CO2 emission per kilometer
You're making two statements here which sound like they're more or less the same, but they aren't. The first is true, the second is not.
An equivalent-power diesel car produces less CO2. But they aren't equivalent. The tax breaks and lower fuel use led to bigger cars. On average Diesel cars have higher CO2 emissions per kilometer.
I don't think that's true (in the UK at least) - both in that it's vanishingly rare to find a petrol vehicle that fit into the lowest 5 UK tax brackets (which are g/km of CO2) - and also in terms of real-world fuel use. I've owned probably 70 cars in the last 10 years or so (I buy and sell as a side gig), and all of the diesels have had lower fuel use for equivalent performance. They tend to be 'one size bigger' in terms of engine (so, say a 1.9 instead of a 1.6, or a 1.4 instead of a 1.1), but even taking that into account they will be a 10-20 mpg better on fuel, with a petrol 1.4 focus giving maybe 30 mpg average, and a diesel (1.6) one easily making 45-50 mpg, more on a run. And that means lower CO2 emissions.
Yeah, you might see the occasional stupidly large diesel car/chelsea tractor, but most of them are 1.9D VAG, and 1.6 or 2.0 Peugeot/Citroen/Ford HDI/TDCIs - there are hundreds of thousands of them out there on the roads.
We don't know that. Diesel is far more prevalent with bigger cars because they tend to eat through more energy over their lifetime, making it more likely to break even on the up-front premium for diesel (not just more joule per kilometre, but typically also far more kilometres per year). But chances are that without diesel, cars in Europe would be just as big today (e.g. they still tend to be smaller that in a certain famously petrol-centric country)
The push should have been for Propane.
Propane is cleaner than almost any other carbon source. Also, you don't have to get it from the Middle East.
Also you should have mentioned LPG, to allow it to be related to existing initiatives (which have now seen decline).
This is a little bit exaggerated. How about the Internet? Or GPS?
Similar to Goodhart's Law - "When a measure becomes a target, it ceases to be a good measure" [1]
That seems even crazier on the surface. Let’s kill a few people so our users don’t have to refill a tank as often.
>after all the scrutiny diesel cars are still being sold and as far as we know pass emissions.
[1] https://en.wikipedia.org/wiki/Worldwide_Harmonised_Light_Veh...
[2] https://en.wikipedia.org/wiki/European_emission_standards#Em...
(here "conformity factor" CF is the relative violation, so "0% conformity factor" means at the original limit, "100% conformity factor" means twice the original limit!)
>In 2017, the European Union will introduce testing in real-world conditions called Real Driving Emissions, using portable emissions measurement systems in addition to laboratory tests.[24] The actual limits will use 110% (CF=2.1) "conformity factor" (the difference between the laboratory test and real-world conditions) in 2017, and 50% (CF=1.5) in 2021 for NO x ,[25] conformity factor for particles number P being left for further study. Environment organizations criticized the decision as insufficient,[26][27] while ACEA mentions it will be extremely difficult for automobile manufacturers to reach such a limit in such short period of time.[28] In 2015 an ADAC study (ordered by ICCT) of 32 Euro 6 cars showed that few complied with on-road emission limits, and LNT/NOx adsorber cars (with about half the market) had the highest emissions.[29] At the end of this study, ICCT was expecting a 100% conformity factor.[30]
EDIT: adding; to me it seems like the regulators knew all along about the defeat mechanisms, and the VW scandal is just the first step: removing plausible deniability of the regulator. Next steps will be when people realise the conformity factor of their car is increasing over time as the laws get stricter? ;) Then people will start seriously questioning the regulators who formally accept such nomenclature!
I don't see any evidence that they "knew all along" but this discussion is very common. We've seen it as well with the financial crisis. A set of people get discovered committing fraud and yet the discussion quickly shifts to how the regulators didn't do enough to stop it. Regulators are often inept and regulatory capture does exist. But if regulators were asleep or even worse complicit does not in any way excuse VW. In fact the complete opposite is true. If the industry somehow subverted regulation then that's an additional crime and not in any way a reason to throw up our hands and say "everyone was doing it, the regulators knew all along, that's life".
Consider a car just barely passing a CF 2.1 NOx test in 2017, then it obviously won't pass the stricter CF 1.5 NOx test in 2021. even though the NOx emissions limit will be identical!:
>European emission standards for passenger cars (Category M)*, g/km
Tier Date (Type Approval) Date (First Registration) CO THC NMHC NO x HC+NO x PM PN [#/km]
Diesel
Euro 6c - September 2018 0.50 - - 0.080 0.170 0.0045 6×1011
Euro 6d-Temp September 2017 September 2019 0.50 - - 0.080 0.170 0.0045 6×1011
Euro 6d January 2020 January 2021 0.50 - - 0.080 0.170 0.0045 6×1011
As the emissions standard becomes stricter, then that same car would only pass increasing Excess Factors
A 2017 NOx Excess Factor currently named "110% conformity factor CF" of 2.1 would result in a "50% conformity factor" car in 2021, even though the emission limits and the car are identical... does that make sense to the reader who wasn't pointed out the inversely defined terminology?
It reminds me of "symbolic euro" which does not mean a euro as a symbolic gesture at all, it means a sum of money representing however high the cost will turn out to be, say in damages,...
found that not one single brand complies with the latest air pollution limits (‘Euro 6’) for diesel cars and vans in real-world driving
The important bit is "real-world driving". As far as we know everyone optimized for the test and yet the driving cycle was not realistic. That is not illegal and supposedly the test has now been improved. What VW was caught doing is much worse than that, it's actually cheating during the test. Emissions testing has been shown to be highly fallible but what VW was caught doing is on another level.
While emissions regulations have a noble goal we would be far better off regulating farm equipment, stationary generators, and boats at this point.
A single farm tractor or lawn mower pollutes as much as hundreds of cars. A container or cruise ship, many thousands of cars.
It reminds me of California approach to water regulations. Squeeze the populace when industry is using 90% of the water.
[0] https://en.wikipedia.org/wiki/Smog#Photochemical_smog [1]https://www.alentecinc.com/papers/NOx/The%20formation%20of%2...
Edit: Worth clarifying that thermal efficiency and combustion temperature are also correlated, which is not plainly stated in [1]
Storing large amounts of cash in bank vaults also creates strong perverse incentives, but we do it and create strong disincentives to practically nullify them.
I’m still one of those who think that what Volkswagen did wasn’t cheating anything. They basically “studied for the test” in order to pas it.
I suspect similar cheats are employed all over the industry. For example my motorcycle has a dead spot which people have been known to bypass the clutch switch to avoid. The reason for the dead spot is that the noise test is carried out at a specific speed and by programming a dead zone in the fuel map they can make it pass the test.
The only real difference between this an the VW emissions scandal is that everyone has to suffer the consequences of cheating the test!
> I suspect similar cheats are employed all over the industry. For example my motorcycle has a dead spot which people have been known to bypass the clutch switch to avoid. The reason for the dead spot is that the noise test is carried out at a specific speed and by programming a dead zone in the fuel map they can make it pass the test.
Can you go into more specifics about this? I used to ride and am embarrassed that I have no idea about this "dead spot" you are referring to.
Example A: "Each model of this type of car must pass Test X with numbers lower than listed in Table Y"
Example B: "The cars must have emissions lower than listed in Table Y. The standard test may be Test X"
If it is written like Example A, then sure, the entire specification is the test, they merely designed to the rule.
But if it's more like Example B, then they are toast. The cars clearly don't ordinarily have emissions lower than Table Y, they are merely engineered to appear to do so during the specific conditions of Test X.
If there is one thing I've learned about the law from the excellent attny who will soon also be my wife, it is that you cannot reason about the law without reading it. Logic and assumptions will not help you -- you have to look at the law itself.
It is actually a lot like software. Our guesses about the behavior might be OK for very limited purposes, but we really don't know jack until we trace it in a debugger or read the source.
(tracing & reading source might be considered analogous to reading the cases and reading the law & regs)
VW did something different. They installed defeat devices that are explicitly illegal. And they knew it as shown in documents and mails. VW also pled guilty in the US already.
So, while both things are on the surface similar they are yet fundamentally different.
And if you still don't know it, every single Diesel engine with this most common Bosch ECU has this cheating SW. The problem is only if you use it or not. 80% do use it. From all European cars only Fiat didn't use it, if I remember correctly. Most Asians did use it also.
You are saying VW installed it. No, Bosch installed it on their devices. And it was written by Audi.
Who's responsible? You still are of the impression that only VW uses this cheating device. Everyone is using it.
Responsible for this device is Bosch.
Audi was a major supplier of Diesel engines for the whole group, especially the larger ones which also ended up in Porsche Cayennes. The smaller Diesel engines came AFAIK from VW. True, the software came from Bosch and can arguably be used for development purposes. The exact role of Bosch in all of this is still not clear.
The only companies found to be using the Bosch supplied software as an illegal cheat device is VW so far. Maybe Mercedes did as well. BMW was, as if now, more or less cleared of illegal wrong doing. They still had to pay fines, so.
Other car makers are, as far as we now today, "only" using the legal loop holes. Still shitty behavior but at least legal. PSA (Peugeot, Citroen, DS and as of last year Opel) are themselves a major Diesel engine supplier and while they paid fines as well, if I remember correctly, they didn't use defeat devices.
EDIT: For clarity, PSA wasn't found to use defeat devices. They had to pay fines in France AFAIK.
But Bosch covered their ass by writing something along "Please note that this is just for testing and cannot be used on production.".
Come on, if your billion-Euro client asks you (or your boss) for this software...
The issue is regulators write regulations for one thing, but want a subtly different thing.
If you and I start a game of poker, but rather than winning through skill, I call the FBI and get them to search all your cards, have I broken the rules of poker? It wasn't a written rule not to call the FBI...
Regulators across the world have similar gaps in the rules, and both humans and automated design tools will start to exploit them, possibly unknowingly. Instead they need to make far more watertight rules. They should simply require all cars to measure their own emissions, and bill the companies for any cars which either mis-measure emissions, or measure more emissions in their lifetime than an allowed limit. Then manufacturers will have an incentive to upgrade cars in the field.
Since when is demand sufficient to bring something into existence?
"When a measure becomes a target, it ceases to be a good measure."
But it was a written rule not to use a defeat device.
This isn’t a case where manufacturers found a loophole and regulators got upset because they wanted manufacturers to obey the spirit of the rules rather than just the letter. This is a case of manufacturers blatantly breaking the explicit written rules.
If I am designing a printer and you say "it must be able to output 10 pages per minute printing black and white text, and make it as light as possible", then I, the printer designer go to work, tweaking the design of the printer to meet that goal, while keeping the weight down.
Seeing nothing about the color printing speed, I don't care about that. It's just black and white speed that matters! I make the black cartridge super fast and super wide. Whoa - gotta keep the weight down! Oh well - I'll make this colour cartridge smaller/cheaper/slower/lighter.
Suddenly, we have a printer that does 10ppm of black and white, and only 0.1ppm of colour, because it has to crawl along.
We now have a 'defeat device'. It performs great in test conditions (black and white text), and badly in general use (colour pages are common).
Did the designer explicitly make a defeat device? Not really... They just only made it work well in the test conditions, and performance in all other conditions suffered.
With your example, you’d be stuck arguing to a judge that color printing is not a normal part of your device’s operation and therefore your loophole exploitation is legal. Which isn’t going to get anywhere at all.
The law is not stupid nor is it evaluated by machines.
Car makers didn't optimize what they were told to optimize for, they optimized for the test they knew the regulators would perform.
A correct analogy would be you telling manufacturers, "the printer must use at most X grams of ink per square centimeter", but then they go and only optimize it for the specific test printouts you will do. In truth their printer is exceeding that threshold just so they can also advertise higher printing speeds.
Even game-theoretically perfect regulation does not work if they're violating the rules.
But I doubt that any popular politician will have the guts to push such a bill against all the industry lobbies :-/
I am so tired of replacing devices just because you can't fix broken software that I see very few alternatives to forcing the manufacturers by law. Just to give an example:
The Samsung Galaxy S3 was shipped with a firmware bug [1] that killed the memory occasionally (just the layout of some internal table so the hardware was completely okay). Sure, when it happened during the warranty period, it wasn't good for Samsung but afterward, you thought your hardware was broken when in fact just the software was broken (and a factory reset doesn't fix firmware internals).
The reason we even know about that bug is that someone disassembled the firmware years after the device had a relevant market share. If the source code for the firmware would have been available, the bug probably would have been made public much earlier and a lot of broken devices would have been repairable.
So I totally understand that manufacturers want to keep some things secret about their products but as a consumer, I feel betrayed when the devices I buy contain time bombs encoded within their software... And since the manufacturers fail to bring recent updates to the devices even during the warranty period, I think it might be a valid option to enable consumers to care for their devices.
[1]: https://media.ccc.de/v/34c3-8784-emmc_hacking_or_how_i_fixed...
I don't dispute that it'd be useful, but the standard for whether it should be statutorily mandated ought to be way higher than that.
>The Samsung Galaxy S3 was shipped with a firmware bug [1] that killed the memory occasionally
I don't really see that as fundamentally dissimilar to a single shoddy physical component bricking an otherwise perfectly functional device.
Consumer devices _used to_ ship with schematics though...
Some (maybe most?) Mercedes units use a Bosch ECU. I'm pretty sure you can dump the firmware from them and figure out how they work fairly easily?
It is however much easier to do some social engineering or hire a bosch employee :). Mercedes, porsche and bosch all have their HQ in Stuttgart, I think it is quite easy for them to know what happens with competitors.
Moreover now with Autosar, they scatter vehicle functionalities through ECUs, with each ECU doing a part of the functionality. It's getting really complex now.
To understand the issue with reverse engineering, one needs to understand that there is no standard in car. Each brand and car has its own communication matrix and electrical architecture, which get periodically overhauled. So even if you manage to identify the meanings of the CAN frames for a make, it is useless for the next generation.
That's why you do not find the communication matrix on internet for recent cars, you only find them somehow for old cars. Also because very few people know how to read CAN network and want to invest timr in this reverse engineering
Excuse me if I sound ignorant but... aren't you able to put the ECUs into a special bootloader mode where you can dump the encryption/protection?
Daimler doesn't do OTA but Tesla for example could remove autopilot code in an update that caused a death and there would be no way to test for it without getting access to the old firmware or source code. Tesla would not hand those out without court order which would require suspicion.
Raid an office, check the source repo and the release log, question engineers, asking questions about what was pushed out OTA, and what wasn't. If your company did not keep such audit and repo logs, you are probably not SOX-compliant (Or the european equivalent thereof), and they should assume the worst in your intent.
Regulators and police aren't some kind of unthawed-from-the-ice-age caveman lawyers. They have as many tools to answer this question as any engineer working at these firms.
The SW is flashed, not OTA. You'll have to do that in service. And the flash update just sets the low temp. limit trigger to 0, disabling the cheat controller.
We know that for years already. The list is public, but no newspaper is printing it. The scandal is only how protected these criminals are. From the affected ~10 carmakers only 3 were punished so far.
[1] https://en.wikipedia.org/wiki/Engine_Manufacturing_Centre
This badly written line is the only technical information in the article. Since a thermostat is only a temperature sensor, I don't get how it constitutes a defeat device of any type. I also don't get how activating a sensor to read coolant temperature protects anything. And why would an installed sensor be turned off anyway?
There is a direct correlation between engine operating temp, condition of the thermostat and emissions/fuel economy.
This is true on gasoline engines, which are homogeneous-charge, air-throttled designs, and target an operating point as close to stoichimetric as possible. Diesel engines are fuel-throttled, and by design run well lean of peak-EGT, stoichimetric, and even peak power due to their stratified-charge nature (overall stoich is rich-as-fuck at the combustion site). A diesel will blowing black smoke and melting holes in its pistons before it ever comes close to stoich, much less rich. A (relatively) richer running diesel engine will be making more power, not less.
A diesel running cold with an open thermostat won't necessarily have it's fuel ratios affected so much as its injection timing. Injection timing absolutely does have an effect on emissions and fuel economy at any one given fueling/power-output level.
Old cars had fans that always ran, often times off the crankshaft. If cold coolant ran through the radiator it wouldn't maintain proper temperature with the fan running! Same situation with highway driving. There's so much airflow to cool the engine - being able to block off coolant is very important. Newer cars with electric fans use this in combination with the thermostat to provide more efficient cooling (by not running fans all the time).
Cars with electric fans use a coolant temp sensor (sometimes on the radiator, or the block, or elsewhere) that kicks over a switch to turn fans on and off.
Then you have a temperature gauge which sends a signal to your dashboard/ecu. The computer uses this in combination with a bunch of other data pieces to make the engine run smoothly.
Anyway, that was all just random info
https://www.sueddeutsche.de/auto/daimler-mercedes-glk-rueckr...
As i understand it, the motor seemed to artificially open the thermostat to keep the engine colder than usual. Now my guess, as an automotive engineer, is that they did this to reduce the time where the engine is in the normal operating range. It makes no sense to me, unless the emissions test only applies when the engine is in normal temperature. Or, second guess, as the engine is colder, it emits less NOx, because NOx appear when the combustion temperature is too high.
Or simpler: the thing activates when the coolant is 20 deg celsius, the starting temperature of the test
That's a major issue. NOx emissions are caused by engine heat interacting with nitrogen and oxygen from outside air, which gets you to a nasty trade-off. Run the engine hotter and get more efficient combustion, therefore less CO2 emissions; or run the engine cooler and get less NOx emissions. By design, Diesel engines combustion happens at much higher pressure and temperature than gasoline engines, which is why diesels produce much more NOx.
Diesel emission fluid can reduce NOx, but also reduces efficiency, meaning more CO2 will be released.
For the article, i think the journalists just didn't understand the explanation, and it is probably the simpler solution : they enabled the emission cleaning algorithm when the coolant at startup is 23 degrees, so clearly a defeating device.
Is this worded correctly? I thought the issue was that the software detected when it was being tested and then output emissions that were substantially less than during normal operation.
This sentence from the article makes it sound like the software made the emissions be under the limit all the time.
The use of the word “pass” leads me to assume that it’s referring to levels when being tested
> ...would PASS emissions limits imposed by the US.
(Emphasis added by me)
I though the type approval was EU wide
Unless, you could somehow prevent competitors from stealing I+D. Which you can't. So closed source it is (that makes things harder for competitors).
This really should have been a requirement for software patents from the start, given how difficult software is to reverse-engineer.
> There should be no room for secret functionality
?
- More humans available to identity bugs and undefined behavior
- Innovation in the form of programmers improving software on their automobiles and sharing their modifications
- Regulators are no longer the only group (other than the maker) with access to source code. This allows new entries into the automotive space without (badly) re-inventing the wheel. This breeds competition, and is a win for consumers.