Did they miss a trick?
<Flight of Fantasy Mode>
Let's say that there are a number of measurable parameters and a number of controllable parameters within the Engine Management System which in combination affect performance and NOx emissions.
Now, if I were to be presented with such a problem I would be inclined to try writing the software and its associated algorithms to goal seek a performance/emissions optimum for any particular set of operating circumstances.
It may be the case that under certain conditions the solution fails specific test requirements but overall the solution averages a performance over the time of a particular driving cycle that meets or exceeds the requirements.
As a result for a 'static' urban cycle test with optimisation switched on the vehicle passes with flying colours but it is possible that under different dynamic conditions with optimisation switched on it fails miserably and yet, as suggested before, a more realistic average brings it back within the required limits.
Then you assume/claim that in as much as the system is goal seeking an optimum for any instantaneous set of conditions it is also adjusting its operating or starting point over time to compensate for things like engine wear or component performance changes such as carbon build up, decline in catalytic converter performance or lubrication performance to maintain optimum performance.
Think along the lines of a neural network whereby it might goal seek a local minima but the true minima is somewhere next door and it needs an occasional kick to find itself in the right place. Again the goal is to achieve an average performance over time that meets the requirements. If you hit it with the wrong specific conditions it fails miserably but overall it passes.
So... you have a system that over time 'learns and adapts' based on instantaneous adjustments and those that occur over time in order to pass the emissions requirements whilst still giving good overall performance in respect of other desirable factors. For example fuel economy and power output and also operates over the lifetime of the vehicle to maintain such performance.
You also do not have to be overly concerned about gee whizz maths to attempt to characterise or determine what the answer should be. You just run the car through a 'bucket load of testing' and let the system get on with doing its job and when it is 'there', or near enough, that becomes your production 'ROM'. Bung it in the rest of the cars and let them get on with tuning themselves.
Of course if your 'bucket load of testing' is largely restricted either by choice or oversight to what the regulatory authorities require, seems highly likely, then your system 'learns' and becomes optimised to those particular circumstances and... all of a sudden, it looks like it is cheating. Also by its very nature you 'do not know' how it does what it does. It just does it.
Having come up with this super whizz idea/system of course the first thing you do not want to do is 'patent' it. Either every other car manufacturer is doing the same or a similar thing or they are not and you certainly would not want to alert them to your methods...
I suppose in the first case you would not be able to get a patent, outside of the US, because it is pretty bloody obvious. After all. I am not 'skilled in the art' but I have managed to come up with the suggestion. Also it is software and effectively based on 'black magic'.
"System that Guesses the Right Answer"??
I might also guess that if you were to give any hint to the authorities as to what your 'game' was that they would suffer a brain
fart short circuit and blibble about demanding that you provide 'proof', 'Turin Complete'?, that this 'splodge' of code actually does what you say it does on its tin but you cannot because you don't know yourselves. It just 'passes the tests'.
As a result it is much easier to let the authorities run the tests or run them yourselves according to the laid out requirements and print out the results demonstrating that it 'works'. Then one day someone wangs it out of the ballpark and discovers that it fails or finds a set of conditions where it fails miserably and starts blowing whistles.
Do you fess up as to what you are really doing and give the 'real' game away to everyone else who is either doing the same or are so far behind the curve because they are not or keep quiet and take a hit to your bottom line whilst pretending you have rogue engineers but you are going to sort things out which may or may not, but is likely to, include providing a bigger urea bottle.
One way you give away your, non patentable, 'secrets' and everyone else gets to benefit from your previously 'secret' technology.
The other way people are happy that they have a bigger bottle of piss in their cars and the supposed problem has supposedly gone away. In the meantime your "System that Guesses the Right Answer" is still in place, discovers that it has a bigger bottle of piss to drink from and learns how to use it.
</Flight of Fantasy Mode>
Of course all of the above falls over if, on analysing the source code, the authorities discover.
While BeingTested(Params.Rec) = True then else GetYourMotorRunning(nil);