HN Explorer

Dieselgate: How exactly did the software know the car was being tested?

7 points by lollipop25 ↗ HN
I just got curious on this Dieselgate scandal. I read that VW used software to check if the car was under test conditions. I know that a car has a bazillion sensors, controllers and such but...

How exactly did the software know it was in a test environment?

I'm still catching up with the story, but I'm more interested on how the software managed to detect the test environment. Is the test environment really so controlled that one could just simply pop in code to tell the car to cheat whenever sensors picked up the right combination?

9 comments

[ 3.2 ms ] story [ 31.9 ms ] thread
JorgeGT
Researcher here, you got it right on your last paragraph. For instance this is the European Cycle that is tested: https://upload.wikimedia.org/wikipedia/commons/thumb/d/dd/Ne... As you can see it is very well defined. Other tricks that are rumored are very simple: you can see the cycle last 20 min. Just instruct the controller (ECU) to change the parameters only after 20 mins of powered state.

In any case, you must know that all those practices were very well known by United States regulators, the term "defeat device" for instance appears in regulation since at least 2007: http://www.gpo.gov/fdsys/pkg/FR-2007-02-26/pdf/E7-2667.pdf#p...

§ 86.1809–10 Prohibition of defeat devices. (a) No new light-duty vehicle, lightduty truck, medium-duty passenger vehicle, or complete heavy-duty vehicle shall be equipped with a defeat device. (b) The Administrator may test or require testing on any vehicle at a designated location, using driving cycles and conditions that may reasonably be expected to be encountered in normal operation and use, for the purposes of investigating a potential defeat device

As you can see it was known that a) defeat devices existed and b) the way to catch them was to try different, random cycles more close to real conditions. The question is, if this was a well known problem and they knew the way to test their existence (and given there were millions of cars on the street with these systems), why start the big scandal precisely now?

By the way, apart from more random, realistic cycles, upcoming European regulation introduces driving tests where the exhaust gases will be collected and compared with the lab test, and only a certain % difference will be permitted.

UnoriginalGuy
> The question is, if this was a well known problem and they knew the way to test their existence (and given there were millions of cars on the street with these systems), why start the big scandal precisely now?

From what I understand the scandal was started after Europe funded research at an American university (since they were developing systems to measure emissions during normal driving that didn't rely on the car's computer output). The results were published a few years ago, which got reported to the US and European authorities, and it is big news right now simply because the fine/punishment phase is in motion (nobody cares about dry research results).

Up until recently (last three-ish years) the technology to measure emissions from the tailpipe during driving weren't readily available. They trusted the computer.

tmuir
I wonder if any hedge funds cashed in on last weeks 30% drop in VW stock as a result of this information being available for the last year and a half.
nvusuvu
I was wondering how to take advantage of situations like this that arise in the future. Be mindful of college level research of corporations.
8thcross
As per what i have read, its a simple combination of front wheels turning, while back wheel at rest plus no steering wheel activity. There may be other factors - but this is good enough to say that car is being tested.
[dead] TerryADavis
You have to talk to God or He won't talk to you. Be witty and charming and never earnest.
stephenr
> Dieselgate

Please don't use this name. Its ridiculous that the media use the "gate" suffix to indicate a scandal of some sort, I'd like to think HN reads/commenters can be slightly more intelligent than the average crowd-pleaser journalist.

creature
I don't think there's any suggestion that VW used these techniques, but there are other indicators that could be used. Things like "If the car door or bonnet is open while the engine is running, switch to low-emissions mode", or using the GPS information (a GPS-measured velocity of 0 with an engine running in gear suggests that the car's on rollers).

I've heard stories of cars using these techniques to cheat in tests, but I couldn't find any reliable citations.

lollipop25
> a GPS-measured velocity of 0 with an engine running in gear suggests that the car's on rollers

Heck! Even if I don't use velocity or gears sensors, the fact that the car is in a testing area (given by GPS) might be enough to switch to low-emissions mode. But I would need the GPS location of every testing location.