By now the full gravity of Volkswagen’s cheating the EPA’s emissions testing systems is out in the open, and while, sure, I’m appalled and disappointed, a perverse part of me is also kind of impressed. It’s a pretty clever cheat! But like all cheaters, VW got caught because they got lazy. I think I know how they could have pulled it off.
Just to recap, the very basic idea behind how VW was gaming the EPA’s diesel emissions-testing system was that they programmed their cars’ engine control computers to check for the specific parameters that would indicate the cars were not on the road, but in an EPA testing situation, which is indoors, on a dyno, and with a very specific set of criteria and actions that need to be followed.
To determine that the car was being tested by the EPA, the VW’s ECU was instructed to check a number of sensors and inputs, including:
• Speed (knowing that the EPA tests dictated certain speeds in a certain order was a big help)
• Steering wheel position (on a dyno, you don’t steer! They just needed to see if the wheel was immobile for unusual periods of time while the car seemed to be in motion)
• Duration of engine operation (the tests had specific and regular timing)
• Barometric pressure (I think this was to check if the car was indoors?)
Again, as amoral as this is, you have to admit, this is a pretty clever idea. When the car determined that these conditions were being met, it switched the ECU into ‘dyno mode,’ which vastly improved emission controls at the expense of performance.
(One question, though — did no one ever dyno these cars just to check the HP? Because it would seem it would kick into the ‘clean’ mode, with less power, whenever it was on a dyno of any kind. We’ll have to look into that.)
So, clearly, whoever at VW came up with this idea was one clever bastard. But, like so many clever bastards, I think he (or she) underestimated the people they were trying to fool. And they got cocky. I mean, why else would the car not default to clean mode? To play things extra-safe, the car should start off in clean mode, and only go to fast/pollute mode when it was determined to be all clear of exhaust snoops.
But the biggest oversight is the one that brought them down: the gig is totally up if anyone tries to do a real-world road test of the cars. So what could the cheater have done to be ready for that?
I think I have some ideas. I mean, it’s too late now for VW, but if any other company is considering a similar cheat, or already doing one (which would not shock me), pay attention.
I think the cheat-designer had the right idea: try and use any inputs from the car to determine if it is in a specific type of location. And, as an aside, I think this may be the first time in automotive history a car has been factory-designed to figure out a location it’s in and change its behavior accordingly. This is a milestone, of a sort.
But the problem is the cheat is only checking a few (admittedly key) criteria, and there’s actually a lot more that could be checked, both to guarantee it’s in a testing facility, and to decide if it’s being road-tested.
I’d check for a lot more stuff: is the radio/entertainment system on and in use? If so, I’d say that’s a vote that you can go into full power, full-pollution mode. The EPA guys probably don’t turn on the radio when they have a car on a dyno.
Also, there’s sensors in the seats to determine if the seat belt buzzer needs to go off — use that, too! If the seat is unoccupied and the engine/wheel speed is over 0, that’s a big vote that the car is on a dyno.
And how about this: VW must know where the EPA’s testing facilities are. These cars — by and large — have GPS navigation, so why not check the GPS coordinates and cross-reference them to a database of where the EPA test facilities are?
Why stop there? These cars almost all have backup cameras now — why not take a picture, and use some image-analysis software to make a good guess at where the car is? If there’s no horizon, and the lighting feels consistent with an indoor context, or there’s non-car equipment right by the camera, I think a reasonable guess can be made that the car is in testing. If autonomous car software can determine moving people from video feeds, a quick, basic image analysis of a frame or two can’t be impossible.
These all help to guarantee if it’s on a dyno, so what about road testing? This gets trickier, but I think is still possible.
First, and most obviously, we know that road emissions testing takes a lot of equipment. Look at these pictures from the WVU report that got VW in trouble:
There’s a whole Honda generator in there, big heavy boxes, computers, and a crapload of piping. All that equipment weighs nearly 700 lbs! Notice also that it doesn’t allow for the trunk lid or hatch to close, in either case.
So, this makes it easy for VW. If the car detects that there trunk or tailgate is open, and if the car senses a huge amount of cargo weight (some cars can detect this; it may require an extra weight sensor or two in the cargo floor of the cars, but that seems like it’d be worth it, now) then the car should stay in the default ‘clean’ mode.
The rear-view camera is often pointed in a non-useful direction when the tailgates are open, but it may not hurt to grab a picture anyway, just to see if any pipes or other obvious equipment is visible.
The ideal situation would be to be able to check for the presence of a tube connected to the exhaust pipe, physically. They could have added a little switch or pressure sensor, but that may be visible to testers, and it would clearly raise some red flags. Perhaps they can check for a back pressure increase that may be associated with a tube on the exhaust? Maybe a sensor further back under the car, near where the exhaust hanger is could check for any strain caused by the pressure of a hose connected to the exhaust pipe? They could say it’s just a sensor to check the integrity of the exhaust hanger system, or something?
And, if we really want to get creepy, how about this: almost all of these cars have an interior microphone, for Bluetooth phone calls and voice command stuff. Hypothetically, the car could turn the mic on and listen for key words only likely to be spoken by people conducting emissions tests: nitrous oxides, chemiluminescence detector, non-dispersive infrared spectrometer, and all the associated acronyms. The sort of emissions-geek crap that no normal human owner is likely to ever utter.
Volkswagen’s secret cheater had a pretty wonderfully devious idea, but, clearly, they didn’t take it far enough. When you’re trying to pull off something this big, this sinister, you really need to overdo it, I think. If VW had been more aggressive in the criteria they use to decide whether or not it was okay to spew more pollutants, and had thought it through enough to at least try to determine if they were on a road test, they might very well still be getting away with it today.
If it hadn’t been for those meddling kids.