By David Miller
The VW Corporation, like their Beetle model, has recently gone “retro”.
VW pulled out a software trick from the 1990’s fuel & emissions playbook – and it is turning out exactly the same way today as it did in 1999.
But first, a little more background…
During the mid-1990’s, the US Environmental Protection Agency (USEPA) had been experimenting with a wonky, clunky little device (they called it the ROVER) that could provide emissions testing on vehicles – out on the road.
At that time, all vehicle emissions testing was done in a laboratory, on a chassis dynamometer, with the vehicle securely locked down. A driver would goose and gun the engine in order to align the cross-hairs with the speed trace of the test cycle. After a certain predictable and repeatable “test procedure” was finished, the next vehicle in the queue could take its turn on the dyno. A simple and inexpensive way to demonstrate compliance.
Which is why the “on-road” thing matters, because – as it turns out – repeatability allows people to “game” a system.
So what is [link style=none url=https://en.wikipedia.org/wiki/Gaming_the_system align=left000 icon=none000 css=none000 color=FFFFFF hovercolor=FFFFFF]”Gaming the System”[/link]? And why does it matter?
Wikipedia defines ‘Gaming’ as “…using the rules and procedures meant to protect a system in order, instead, to manipulate the system for a desired outcome.”
Therefore, gaming requires repeatability and a lack of randomness in order to succeed. Unfortunately, most governmental and regulatory entities require just that: predictability, repeatability, steadiness, transferability, standardization – order. Historically, this has been a good strategy, and one that has been economically viable.
However, even “house advantage” economic viability may be a myth of sorts; consider[link style=none url=# align=none000 icon=none000 css=none000 color=FFFFFF hovercolor=FFFFFF]Card-Counting[/link]as one example.
One way of gaming a system in cards is to assign positive, negative, or zero values to each card. When that particular card value is dealt, the “card count” is adjusted by the assigned counting value.
Low cards increase the count as well as the percentage of high cards in the shoe, whereas high cards decrease the count. Card Counting’s goal is to assign point values to a card’s “Effect of Removal” (EOR). The EOR is the impact on the house advantage, which allows the Card Counter to calculate their strategy; one typically utilized mathematical strategy often employed with Card Counting is called the “Kelly criterion”.
The [link style=none url=https://en.wikipedia.org/wiki/Kelly_criterion align=none000 icon=none000 css=none000 color=FFFFFF hovercolor=FFFFFF]”Kelly criterion”[/link] encourages an increased bet according to the proportionality of the Card Counter’s advantage. The higher the count, more is bet on each hand – which leverages the Card Counter’s lead.
We have been here before – leveraging and counting cards, so-to-speak. And it proved to be a costly gambit.
Back in the 1990’s, several large and well-known manufacturers (OEMs) figured out the predictability and repeatability of federal testing procedures (FTP’s) utilized by the US Environmental Protection Agency (USEPA). “Defeat Devices” (in the form of engine control software) were installed, so that engines could increase fuel efficiency at the expense of higher NOx and ultra-fine particulates on road, while switching back to an emissions-friendly operation when on an engine dynamometer test. It was an easy fix for the OEMs. The onboard software looked for predicable patterns of acceleration and deceleration – ones that matched various FTPs – and quickly made the switch.
It worked well, until after multiple warnings from whistle blowers about the cheating, the USEPA finally tested an engine by simulating steady-state, highway operation instead of the certification test cycle; huge difference. One that led to the billion-dollar [link style=none url=https://en.wikipedia.org/wiki/Consent_decree align=none000 icon=none000 css=none000 color=FFFFFF hovercolor=FFFFFF]Consent Decree[/link] OEM fine of 1999.
What did the Consent decree of 1999 change?
For starters, it funded a ten-year run (2000 – 2010) of “Portable Emissions Measurement Systems” (PEMS) development. The Consent Decree also led to new field standards such as the “Not-To-Exceed” (NTE) Standard, as well as some of the impetus behind more rapid identification & deployment of newer, more effective technologies (e.g. Diesel Particulate Filters (DPFs)).
However, these gains and lessons-learned in the Heavy Duty Diesel (HDD) market have apparently gone ignored in the Light Duty Diesel (LDD) market. Is it because LDD OEMs were ignorant of the industry changes?
No – it is simply that there have been no easy methods to introduce randomness into the regulatory process.
And VW (and possibly other LDD OEMs) have been able to leverage the rules and procedures meant to protect a system, instead, to manipulate the system for a desired outcome.
The real Card Counting challenge isn’t about meeting – or beating – a governmentally-implemented metric. It is about meeting the very serious challenge of reducing fuel usage and emissions reduction.
Perhaps the “Kelly Criterion” should be more carefully considered: although it may initially allude to a better strategy in the long run, built-in constraints may “override the desire” for a more optimal – and realistic – growth rate. In other words, the desire to cheat, based on the knowledge of patterns, becomes the Card Counter’s demise.
In this particular case, VW did not consider the long game.
What’s next? Will the USEPA impose a Part II Consent Decree? Will VW be part of a funding mechanism for “next-gen” integrated PEMS (iPEMS)?
Stay tuned!
______________________
[1] Michael Walsh, “Asleep at the Wheel; The Environmental Protection Agency’s Failure to Enforce Pollution Standards for Heavy-Duty Diesel Trucks – A staff report, prepared for use of the U.S. House of Representatives Committee on Commerce.” March 2000.
As suspected; see the following:
http://www.thefiscaltimes.com/2015/09/22/Volkswagen-May-Not-Be-Only-Car-Company-Cheating-Emissions
Perhaps PEMS with dash board display of emission levels reached on actual running of the vehicle would be a better deterrent