Grex Agora46 Conference

Item 66: Car air bag "black box" used as evidence

Entered by gull on Mon Jul 7 15:42:49 2003:

I figured it was only a matter of time until this happened.  I can't say
I can think of anything really wrong with it right now.


Evidence from a car air bag's electronic data recorder played a vital
part in sending a drunk driver who killed two girls to prison for 30 years.

Edwin Matos, 47, was drunk and speeding in a suburb near Fort
Lauderdale, Florida when his Matos' 2002 Pontiac Grand Am ploughed into
a car driven by a teenage girl who pulled out of a driveway into his path.

Jamie Maier, 16, and her friend, Paige Kupperman, 17, were both killed
instantly by the resulting crash last August. Matos survived.

In court last month, Matos's lawyers claimed he was travelling 60 mph in
a 30 mph zone. But data from the air bag's electronic data recorder
showed he was travelling at 114 mph seconds before the crash.

This evidence contributed to Circuit Judge James Cohn decision to
sentence Matos to a maximum 30 years imprisonment when he was convicted
on two counts of manslaughter for his crimes. The Judge also ordered
Matos to pay more than $17,000 to the two girl's families for funeral
expenses, AP reports.

Data recorders in air bags can record a car's speed and deceleration and
other data such as the pressure on a brake pedal at the time of a crash.

An estimated 10 million vehicles in the US are fitted with such
recorders, which vehicle manufacturers began putting in vehicles in the
1990s to test air bag performance.

In recent year court prosecutors have begun using information from these
data recorders, the existence of which most drivers will be unaware, in
criminal prosecutions, AP reports.

(Source: http://www.theregister.co.uk/content/55/31564.html)
36 responses total.

#1 of 36 by russ on Tue Jul 8 01:33:33 2003:

Re #0:  The 114 MPH figure seems rather unlikely; it would take most
vehicles far longer than the length of a neighborhood to get to that
speed, and the idea of the driver of a car surviving a collision at
such speeds is incredible.  Far more likely is that the reading was
faulty, either due to a bad sensor or wheel spin; I bet I could get
a car to read 114 MPH without going over 30, on dry pavement to boot.

Not to excuse the convict, but basing the length of his sentence on the
unvetted "testimony" of a piece of hardware not certified for forensic
purposes seems more than slightly wrong.  Garbage In, Garbage Out; if
the typical sentence for a drunk driver who kills someone isn't also
the maximum, I don't see that Matos should have gotten it just because
of a number in a flash EEPROM.

I hope that the use of such data taken by non-certified systems is
appealed, because either the reliability should be assured by testing
and comprehensive self-checks, or it should not be admissible in court.

A question for anyone who disagrees with me:  would YOU want to go away
for 30 years because of someone's buggy code?  Remember that much of
that code is written by programmers considerably junior (in both skills
and experience) to many of the people here, and the purpose for which
the system is designed is quite a bit different from this application.


#2 of 36 by jmsaul on Tue Jul 8 02:08:45 2003:

I agree with Russ on this one.  I have real concerns about admitting those
readings as evidence, given that the equipment wasn't designed for that level
of reliability and accuracy, and we don't know what its failure modes are.
Or how reliable it is.


#3 of 36 by johnnie on Tue Jul 8 02:31:45 2003:

>The 114 MPH figure seems rather unlikely; it would take most
>vehicles far longer than the length of a neighborhood to get to that
>speed,

Nonsense; I can think of any number of roads where a driver could wind 
it out to 114 before being dumped into a neighborhood with plenty of 
driveways and a 30mph speed limit.  I live on such a road.

>the idea of the driver of a car surviving a collision at
>such speeds is incredible.  

Incredible, perhaps, but certainly not improbable.  Most people won't 
survive a 100mph+ crash, but some do.  



#4 of 36 by senna on Tue Jul 8 03:04:44 2003:

The dead bodies cushioned the blow.  Actually, one would have to know the
details of what physically happened in the accident to know whether or not
survival of such a crash was realistic... but I think it's certainly possible.

Black box technology has been well developed over the years both for use in
aircraft and, more recently, in automobiles.  Black boxes equipped in race
vehicles provide, for the most part, extremely accurate readings of things
like speed, G-force, tire pressure, and so on.  They have to--the amount of
money tied up in what the black boxes say is actually quite high.  I have no
doubt that solidly developed black boxes can provide strong, accurate pictures
of what was happening to a car prior to a crash.

However, lie detectors do a good job of telling whether or not people lie,
too, and black boxes need to be well proven before they should be used as
court evidence.  I have no idea what the pedigree of the black box in the car
involved was.  Any word on whether or not it was well-proven?


#5 of 36 by rcurl on Tue Jul 8 05:41:18 2003:

At what point in the drive train is the speed recorded in the "black box"?


#6 of 36 by scott on Tue Jul 8 06:46:11 2003:

An inertial accelerometer could be used to measure the speed - they've almost
certainly got one in there just to help figure out when the airbag ought to
be triggered.

And hey, "airbags save lives", maybe that's how the drunk lived through the
crash.  Anyway, his real crime was driving drunk, and a simple & reliable test
for blood alcohol would prove that.


#7 of 36 by pvn on Tue Jul 8 08:06:38 2003:

I too agree with both russ and jmsaul on this, but only to a point. 
Sure this is new technology but so were fingerprints and blood typing at
one point but they are pretty much accepted today.  I also don't have
much of a problem with a drunk driver surviving a head on collision with
the side of a vehicle while the passengers of the vehicle hit on the
side are killed to to the lateral acceleration if they didn't have the
proper side impact airbags.  And I suspect that no matter how you look
at it the drunk's black box probably recorded something closer to 114mph
than 60mph in a 30mph zone - ok, so he was only driving twice the legal
speed instead of almost 3 times?  He deserved what he got.

On a side note, one of the neat things you can do with a piece of
surplus USAF test gear that was originally designed to test missle
warning systems on aircraft is to fire off the airbag systems of certain
mercedes and other high cost automobiles.  Originally it was fun to sit
off the active runway and fire them suckers to see what the wingnuts
would do.  THe seasoned would figure "NFW" but the FNGs would do some
really excellent aerobatics while dirty.  So, with the proper
electronics one may bloody the nose of someone parking, but how likely
is that in real life or in this case?


#8 of 36 by gull on Tue Jul 8 14:13:59 2003:

Re #1: The wheelspin argument hadn't occurred to me because I assumed
the 114 mph figure came from extrapolating from accellerometer readings,
since the airbag control unit would have to contain an accellerometer to
begin with.

What I really don't understand is why it matters whether he was going 60
mph or 114 mph in a 30 mph zone.  Either way it's a ridiculous speed to
be going there.

Re #3: How likely he would be to survive a crash at 114 mph depends
entirely on what he hit.  It's not the speed that kills you, it's the
deceleration.

Re #7: A friend of mine has a neat toy...a small Gunn diode transmitter
and feedhorn connected to a 9V battery and a push button.  It's amusing
to hit the button as someone streaks by you on the freeway and watch
them hit the brakes as their radar detector goes nuts.


#9 of 36 by janc on Tue Jul 8 16:17:39 2003:

The objection Russ raises are obvious enough that I'd expect any vaguely
competent defense attorney to raise them too.

I'd think that there would be a lot of evidence that would tend to distinguish
between a 60mph crash and a 114mph crash.  You'd expect that an expert
examining the vehicles and the crash site would be able to make a pretty good
guess about how fast the driver was going, and that you could present other
evidence that would tend to suggest the car was going faster or slower, though
it might not give an exact value like 114mph.  In fact, I would expect that
nobody would bother to check the airbag controller unless they thought the
reported speed of 60mph didn't make sense.  If I was on a jury, it'd be hard
to convince me to convict on the basis of an airbag accelerometer without
some substantial collaborating evidence.  I bet it existed in this case.

You certainly could survive a 114mph crash in the right conditions.  Assume
the victims were in a lighter car without side impact airbags, that maybe was
already starting to turn after pulling out of the driveway, so it was a 
partially glancing blow.  Many modern cars protect much better against
front impacts than side impacts.


#10 of 36 by rcurl on Tue Jul 8 17:35:42 2003:

I did some web searching on how the "black box" works. It has an
accelerometer that triggers the airbags and also writes to memory data for
car speed, engine speed, brake actions, and other things. However all of
these data are from 5 second before the activation of the recording. A lot
can change in 5 seconds. What I was unable to find was how the vehicle
speed is sensed. It is not by means of the accelerometer, as far as I can
tell. 



#11 of 36 by scg on Tue Jul 8 19:12:26 2003:

In my high school physics class, the Ann Arbor Police accident investigation
people closed the street in front of the school one day and did a recreation
of a fictional accident for us to measure.  They taught us how they measured
the speed of a car prior to an accident, as well as where the driver began
trying to stop, by measuring skid marks left over from the accident, and then
taking a car with the same mass and the same tires and skidding it in the same
place.

If the driver were really drunk I suppose there might be no skid marks, but
that would presumably make the physics calculations easier.  Rather than doing
the full recreation, they could just calculate how fast an object of a given
mass would have to be going to throw another object of a given mass a given
distance.


#12 of 36 by russ on Wed Jul 9 02:12:51 2003:

Re #4:  I doubt very much that you will find significant commonalities
between air-transport "black boxes" (which are actually painted orange)
and the automotive devices which bear the same label.  They are designed
for completely different purposes, and are vetted to radically different
standards (if the latter are vetted to any standard at all).

Re #5:  Given that many of the vehicles so equipped probably do not
have the per-wheel sensors used by antilock brakes, the only thing that
the designer of a generic "black box" could use is the vehicle speed
sensor at the transmission output shaft.  Wheel spin can make the
reading of such a sensor be radically wrong.

Re #6:  Sorry, Scott, but you haven't given that enough thought.  The
accelerations of interest to an air-bag trigger accelerometer are
orders of magnitude larger than the rates at which a car can accelerate
under its own power, and the quantization error and zero-point drift
of the output would make it useless to try to integrate the value to
measure speed.

Re #8:  Well, yes, 60 MPH is a ridiculous speed to be going in a
residential area.  The real question is this:  is it just to give one
drunk 30 years for killing two girls, and another one 10 years, because
one of them had a number in an EEPROM and the other didn't?  Somehow I
do not feel safer as a member of the public if chronic drunks start
driving older cars to avoid electronic finks and harsher sentences.
(Hmmm... who just replaced an old Lincoln?  With what?)

Re #9:  The reliability of most other circumstantial evidence does
not depend on un-vetted software taking irreproducible measurements,
and it's all but certain that the defense didn't have the money to
subpoena the code and hire special masters to go over it for bugs.

For more worrisome things, see the items which follow.


#13 of 36 by senna on Wed Jul 9 03:44:30 2003:

Depends on which black box you're talking about.  Typical passenger car black
boxes?  Probably, yes.  The racing versions are developed by teams like
Ferrari with nine-figure annual budget devoted to a total of 17 two-hour
races--those are considerably more advanced and have to deal with many of the
same issues, though what they want to know has different functions.


#14 of 36 by scott on Wed Jul 9 06:31:21 2003:

Russ, you're making assumptions about the capabilities of accelerometers. 
You're also making assumptions about what standards are set for airbag
systems... unless you've actually got some airbag design experience your
assumptions aren't any more valuable than mine.


#15 of 36 by gull on Wed Jul 9 13:39:54 2003:

From what I've heard, car companies have been adding more monitoring
functions to airbag black boxes in recent years.  They're trying to
collect data that will help them improve safety.  I'm assuming, if they
want their data to have any validity, they've thought about the issues
Russ is talking about.  I'm also assuming that a defense attorney would
have thought of the same arguments, and that the black box data was
either confirmed by other observations (examination of the crash scene,
"speedometer slap", etc.) or concerns about its accuracy were addressed.

I heard another news story today that gave a different figure than the
114 mph one, but I don't remember what it was.


#16 of 36 by russ on Thu Jul 10 04:02:21 2003:

Re #14:  An assumption borne of broad experience in the industry, Scott.
An engineering manager will try to squeeze out fractional cent savings
per unit, because a tenth of a cent over the annual volume of a single
car platform is several times his bonus.  This means getting rid of
any capabilities that aren't absolutely necessary for the designated
purpose of the system; if you need to trigger the airbag between 5.0
and 5.3 G of longitudinal deceleration, you can bet that the system
will not be designed to sense and integrate 0.01-0.1 G accelerations
and may very well not be accurate enough to do anything so sensitive.
Placement of sensors in vibration-prone areas may degrade accuracy further.

If you want to hear an interesting story about eking out cost savings in
automotive, pony up a beer and ask me about the Ford Corporate Switch.


#17 of 36 by scott on Thu Jul 10 07:14:49 2003:

So you don't have specific experience with airbag design, then.  


#18 of 36 by gull on Thu Jul 10 13:15:29 2003:

I've always relied on fractional-cent savings to explain the more
bizarre design shortcomings in cars I've owned.  Like the Ford van that
had no coolant recovery bottle, just a hose pointing at the ground.


#19 of 36 by russ on Thu Jul 10 21:59:31 2003:

Re #17:  Are you arguing that you know the business better than I do?

I got put to work on some software design specs once, but got pulled
onto another project before I could get into the meat.  I have more
experience with stability enhancement systems, which do use
accelerometers of much higher sensitivity.  However, SES is only
available as part of very high-end ABS systems; your generic "black
box" is not going to have such data available.


#20 of 36 by tod on Fri Jul 11 00:02:50 2003:

This response has been erased.



#21 of 36 by jor on Fri Jul 11 00:49:55 2003:

        Fix Or Repair Daily
        Found On Road Dead
        First On Race Day

        backwards?
        Driver Returns On Foot

        I wonder if I may be acquainted with Mr. Matos.
        A fellow with the same name contacted me in early
        2001 from a programmer contracting office in
        Boston. He ended up placing me at a job
        assignment for 6 months. It got extended to 11.

        The company is HQ'd in Fort Lauderdale and this
        Edwin Matos ceased being employed there . . 
        around last August, maybe I can look up the
        time frame. Strange thing, for emailing,
        he was in the habit of using my grex address.   

        I guess I shouldn't be surprised to not find
        his employer's name in the news stories.



#22 of 36 by russ on Fri Jul 11 03:58:34 2003:

Re #18:  That used to be standard on everything.  So were vacuum motors
to run the windshield wipers.  Coolant bottles and electric motors are
the modern improvements (some mandated by law, some not).


#23 of 36 by scott on Fri Jul 11 06:36:58 2003:

Re Russ's #19:
Russ, I'm arguing that we're both making assumptions, and that unless you've
got ACTUAL experience in airbag design your arguments don't trump mine.  OK,
so now we know you've worked with accelerometers.  So have I, at MSU's
Agricultural Engineering dept (employee, not student).


#24 of 36 by russ on Fri Jul 11 12:07:15 2003:

Re #20:  No, Ford was neither supplier nor customer for that project.


#25 of 36 by gull on Fri Jul 11 14:05:43 2003:

Re #22: I realize coolant recovery bottles are modern improvements.  But
I considered a 1990 model year van to be "modern".  My '85 Crown
Victoria had a recovery bottle, as does my '82 Volkswagen.

Eventually I got sick of it pissing coolant onto the ground on hot days
and bought an aftermarket bottle.


#26 of 36 by russ on Fri Jul 11 22:28:29 2003:

Re #23:  If you have no experience with the cost-driven nature of
automotive design practices, you don't really grasp the issue.
Claiming that someone else's experience isn't exactly the matter
at issue (therefore you are correct) is special pleading; air bags
are made by suppliers of broad lines of automotive components, all
product lines are under identical cost pressures, and they swap
engineers and managers all the time.


#27 of 36 by scott on Sat Jul 12 06:13:27 2003:

Russ, unless you've got SPECIFIC EXPERIENCE in airbag design your assumptions
are NO BETTER THAN MINE.  I realize you've probably got a lot of emotional
investment in "I'm smarter than everybody esle" but you really need to accept
that it might not be true.


#28 of 36 by russ on Sat Jul 12 06:27:15 2003:

Re #25:  Lots of improvements mandated for cars were not mandated
for "trucks", and vans are legally "light trucks".  I don't know why
a van owner wouldn't want to keep free oxygen out of the cooling system
and the other benefits that a coolant recovery bottle provides, but it's
possible that the space constraints under the short nose of the typical
van (plus the cost) kept the manufacturers from offering them for a few
years.  Hey, the radiator replacement comes after the warranty expires,
so why should they care?  </cynicism>


#29 of 36 by russ on Sat Jul 12 13:57:33 2003:

(Cluelessness:  Using an argument immediately after being told
exactly why it is fallacious.)


#30 of 36 by scott on Sat Jul 12 15:48:30 2003:

(Obstinance:  Pretending to be an expert, even after having to admit one has
no special knowledge or experience in the relevant field)


#31 of 36 by russ on Sun Jul 13 04:02:20 2003:

Re #30:  Yeah, but when are you going to admit it to yourself?

Just for the audience, the issue is whether the airbag-trigger
accelerometers are used to feed data to the "black box" used to
get information on airbag performance.  Scott is claiming that
a system which is absolutely required to produce a binary fire/
no fire (or a trinary no fire/slow/fast) indication at an
acceleration of several gravities will either be suited to
measuring and integrating accelerations of a small fraction of
a gravity, or have its design broadened to accomodate the
additional functionality (at some cost in better sensors, more
computing power, and more complex and thus expensive testing
requirements).   That's what it would take to get meaningful
information during the period before the actual collision.

I'm asserting that this is contrary to automaker design practice
in all regimes, including safety-critical systems such as brakes.
(BT, DT, GTTS.)  The automaker's preferred method of capturing
speed data would be to read the speed-sensor packets going over
the vehicle bus, at zero cost; the bus interface is required to
capture other data, and is far more standard across platforms.


#32 of 36 by scott on Sun Jul 13 05:45:53 2003:

Russ, you're making bad assumptions about my position.

I'm claiming that it's possible, not that the automakers would spend extra
money on a complete implementation.

Although it's also possible that the automakers, for CYA purposes, indeed
spend extra to have accurate acceleration/speed data.


#33 of 36 by pvn on Sun Jul 13 08:49:15 2003:

Hmm.  I do know that it is possible given knowlege of the make and model
to trigger the airbag of many a parked car using a 5lb sledge
appropriately applied.  Think thunder storms and car burglar alarms for
another and similar example.


#34 of 36 by gull on Mon Jul 14 15:24:53 2003:

Re #33: I'm not sure car alarms are a good example.  The culprit in a lot of
cases seems to be "window break detectors" which are really just "loud noise
detectors".  For some reason (maybe so the owner feels like the alarm is
"doing something") these suckers seem to get cranked up as far as they'll go
when the alarm is installed.


#35 of 36 by russ on Mon Jul 14 22:12:32 2003:

Re #32:  The automakers are interested in statistics, which wouldn't
be hurt by a substantial fraction of data known to be unreliable and
thus discarded.  (How many hundred air bags go off every day?)

Forensic purposes require accurate data every time; it is an essential
conflict between the requirements of the base system and the very
different uses to which its output is put.


#36 of 36 by scg on Tue Jul 15 03:34:35 2003:

I don't think Scott is claiming to be an expert.  Scott is claiming that
neither he nor Russ are experts.


There are no more items selected.

You have several choices: