Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

A year of digging through code yields “smoking gun” on VW, Fiat diesel cheats

arstechnica.com

Post reply on HN
Print view View original

141–150 of 151 posts

Re: A year of digging through code yields “smoking gun” on VW, Fiat diesel cheats

#141

Post by vinceguidry »

javajosh wrote: post #111
vinceguidry wrote: post #98

Earlier quoted context omitted.

I'm not waiting for anything. I'm simply recognizing that "encouraging" engineers to be heroes is the lazy way out and is not a real solution to anything. This is a political problem, it can only be met by an appropriately collective effort. > Let's face it, there are other programming jobs out there. OK, let's run with that for a moment. Where else would you have expected the VW whistleblowers to get jobs? Not every…

Maybe, maybe not. This is speculation, unless you've done a study to sample a statistically significant portion of the population. I agree, your speculation sounds reasonable. But neither of us knows , and both of us have certain biases that this speculation confirms - and therefore it should be mistrusted until confirmed. Frankly, I think there is a bigger problem than job security itself. It's that the whistle will…

Everything is speculation until someone makes it real.

Re: A year of digging through code yields “smoking gun” on VW, Fiat diesel cheats

#142

Post by mcguire »

Sunset wrote: post #49
mabbo wrote: post #9

While I certainly don't support the 'rogue engineer' excuse that VW originally gave everyone, the idea does bring up an interesting point. An engineer did write this code. Almost certainly, many of them, working together. Now, their managers told them it was okay to do this, maybe that legal had given the go-ahead or that they had to do it to keep their jobs... But it's still pretty clear that this was not right. The…

>In my opinion (and I welcome disagreement and debate) engineers have an obligation to say 'no' to wiring unethical code. I respectfully disagree. Engineers have an obligation to do whatever is asked of them by management. If that means building nuke-hand-grenades so be it. Being an engineer( be it software or mechanics ) should be a morally neutral thing. Do the job and keep your morals out of it. This is the only w…

Real engineering disagrees: http://www.onlineethics.org/2959.aspx

Re: A year of digging through code yields “smoking gun” on VW, Fiat diesel cheats

#143

Post by mcguire »

denzil_correa wrote: post #72
bambax wrote: post #48

> In 2015, regulators realized that diesel Volkswagens and Audis were emitting several times the legal limit of nitrogen oxides (NOx) during real-world driving tests. But one problem regulators confronted was that they couldn’t point to specific code that allowed the cars to do this. They could prove the symptom (high emissions on the road), but they didn’t have concrete evidence of the cause (code that circumvented…

Someone finally said it. Basically, the test is to be blamed too. Like any other test - you should randomly pick up cars from the road and test for emissions. I really don't see why the regulator can't do something like it.

That way, we can complain about their tests being unscientific.

Re: A year of digging through code yields “smoking gun” on VW, Fiat diesel cheats

#144

Post by mcguire »

adrianN wrote: post #6
tyingq wrote: post #3

Earlier quoted context omitted.

The real world sanity checks for cheating make sense. But the nature of the regular tests can't be secret. Carmakers need to know the bar they are supposed to hit. Building something on the speculation that it might pass, maybe, is a whole different business model.

"Build a car that doesn't emit more than X of exhaust Y when a typical driver drives it a) ten kilometers in the city, b) a hundred kilometers on the highway." That seems like a good enough requirement to me. Define "a typical driver" as "out of a hundred random test drivers, no more than 20 exceed the thresholds, no more than 5 exceed the thresholds by more than a factor two." That forces the car manufacturers to ha…

Are you a software developer? Could you write the code for an application specified that loosely?

(At this point, every one of us is saying, yeah, I have. It just took a long time, a lot of change requests, and many round trips with a customer. Which is the point.)

Re: A year of digging through code yields “smoking gun” on VW, Fiat diesel cheats

#145

Post by mcguire »

wereHamster wrote: post #62
tyingq wrote: post #23

Earlier quoted context omitted.

Driving style has a profound effect on emissions. As does ambient temperature, uphill vs downhill, time at stoplight. Short trip vs longer (catalytic converters don't work till they are hot). I'm sure they can find some way to address the cheating, but it's likely not the case that they are just missing some obvious simple solution.

Modern cars have enough sensors that they could store all necessary information about trips, have the service center read it out each time the car is serviced, anonymise, and send back to the manufacturer. That way the manufacturer would have accurate, real-world data about their 'average driver'. Could even make it country-specific.

Why anonymize it? Could be a new profit center.

Re: A year of digging through code yields “smoking gun” on VW, Fiat diesel cheats

#146

Post by mikeash »

fjdlwlv wrote: post #128
indolering wrote: post #76

Earlier quoted context omitted.

This is also how we end up with "crazy" regulations. Case-in-point: nuclear waste handling. Engineers create complex six-sigma safety plans which are then slowly eroded as the regulations and safety protocols are wacky overkill, right? The WIPP nuclear isolation site had a 15 year run so management began to cut corners. Then three unlikely events all lined up and nearly got people killed. It started with a truck catc…

Seems like the "organic" fad was the problem

The problem was substituting an unsuitable material because of mistakes when revising procedures and insufficient review of the revisions. The fact that the mistake involved an "organic" product is coincidental.

Re: A year of digging through code yields “smoking gun” on VW, Fiat diesel cheats

#147

Post by mcguire »

parennoob wrote: post #10

> In 2015, regulators realized that diesel Volkswagens and Audis were emitting several times the legal limit of nitrogen oxides (NOx) during real-world driving tests. This means that they can detect emissions levels during real world driving tests. What's wrong with just making those tests the actual regulatory ones? So whatever ingenuity automakers can use will be put to minimizing emissions in the exact same scenar…

Can you make those real world tests repeatable without increasing the costs by several orders of magnitude?

Back up real world tests are a very good idea, but the legal limits need to be defined under standardized, laboratory conditions.

Re: A year of digging through code yields “smoking gun” on VW, Fiat diesel cheats

#148

Post by luca_ing »

thyselius wrote: post #66
throwaway79790 wrote: post #59

Earlier quoted context omitted.

The problem is that nobody wrote a specific defeat device. You have to understand, Bosch has one basic (extremely complex) piece of software that goes in pretty much every single engine controller. There are some specific functions that are developed at the request of specific customers, but the basics are always the same. When this general piece of software is set up to control a new engine from an OEM, a multi-year…

I don't doubt it's this big mess you are describing, but if the engine is actively looking for test/not test and does things like turning off particle filtering after 30 minutes of driving because it knows any test will be over, it seems to me the intention is in fact to lie about the real emissions.

The thing is that detecting test situations is good, certainly useful, perhaps necessary.

E.g. detecting you are on a test stand, you might

* deactivate plausibility checks which would normally limit engine power * deactivate air bags and other inherently dangerous systems * ensure correct system behaviour even if sensors give conflicting information * suppress error logs for certain failed plausibility checks

The only illegal thing is switching to a "massaged" parameter map under these conditions. And presumably a single person controlled this switch.

Re: A year of digging through code yields “smoking gun” on VW, Fiat diesel cheats

#149

Post by mcguire »

edejong wrote: post #65

An interesting twist to this would be to change liability. Make the driver liable (not the manufacturer) for driving a car breaking the emission standards. This will set in motion the following steps: - Car users will be offered insurance against breaking emission standards - Insurance companies will hire specialists to lower insurance rates of above insurance - Consumers are disincentivized to modify their ECU in a…

- Create a market for road side emissions testing equipment. Your traffic stops just got a bit more thorough.

- Reduce emissions standards to levels that can be tested easily in nonstandard conditions.

Re: A year of digging through code yields “smoking gun” on VW, Fiat diesel cheats

#150

Post by tetha »

bambax wrote: post #48

> In 2015, regulators realized that diesel Volkswagens and Audis were emitting several times the legal limit of nitrogen oxides (NOx) during real-world driving tests. But one problem regulators confronted was that they couldn’t point to specific code that allowed the cars to do this. They could prove the symptom (high emissions on the road), but they didn’t have concrete evidence of the cause (code that circumvented…

> Or, modern technology should allow to test a car all the time and report emissions and fuel efficiency, etc. during its lifetime.

This can become a privacy issue though. Suddenly, you're able to associate licenses with a set of persons with some probability. And you're able to collect side channel information about the vehicle being in motion, and the speed of the vehicle. And who knows what more you can read from the sensors of a modern vehicle. I know of a couple of vehicles with GPS sensors.

Overall, this discussion is very interesting to read, because this is the discussion about unit testing to a dot. We are unit-testing cars with a mock road - and unit testing fails with malicious and/or stupid workers building the unit. Every set of unit tests can be satisfied by a lookup table - this is happening right now in cars.

So now, the question is: Which of the bigger system control tools do we use. integration tests, so, taking cars on a race track? Property based tests - randomize the length of the test, define ranges of acceptable pollution. Live system monitoring? Maybe a pipeline of tests?

Post reply on HN