Did you know that the airbag control module in
your car could be logging a whole lot of factors about your driving - including
your speed? If the proliferation of speed cameras, redlight cameras and radars
isn’t enough to make you drive carefully, perhaps that piece of news just might!
Think about it. If in a moment of inattention you
collide with the back of another car, you won’t be able to say that you were
braking hard – not if the electronic record shows that in fact you never even
started to slow until the time of impact. Convicted by your car? – it’s more
than just a possibility, with one such case having already occurred in the US.
There a driver involved in a double fatality claimed he had been travelling at
about 100 km/h. But the electronic record showed that in fact the speed of the
car five seconds before impact was 184 km/h...
So what data is logged and why is it recorded? Do
all airbag-equipped cars have this facility? How can you read it? And who owns
The implications - not only for drivers but also
for insurance companies, the police, rental car companies and fleet owners - are
But if the thought of your car logging your
driving behaviour horrifies you, here’s a let-off. At this stage General Motors
in the US appears to be the only car company wholeheartedly embracing the
technology, publicly releasing details on their systems and also working with a
third party provider to make available for general purchase a dedicated data
reader. However, the benefits of Event Data Logging to crash researchers mean
that the US Government is strongly supporting the development of universal
standards and techniques of implementation. In other words, with the influence
that US legislators have on world car developments, it’s probably only a matter
of time before all cars have Event Data Logging in a format that allows easy
About 25 years ago the fuel and ignition control
technologies in cars started a move from mechanical systems (carburettors and
points) to electronic systems (EFI and electronically controlled ignition).
These systems use sensors to measure various car parameters, such an engine
airflow, engine speed and throttle position, with an Electronic Control Unit
(ECU) then making decisions about fuel injection pulse width and ignition
timing. Additionally, most of these systems have the ability to detect and store
faults, allowing their later diagnostic reading.
The increasing sophistication of automotive
electronic systems has led to the adoption of ABS, Traction Control, Stability
Control and Climate Control (amongst others), with each of these systems also
able to log fault codes. Additionally, many car systems share sensors – for
example, the road speed sensor is read by the engine management system, cruise
control system and traction control system. This ‘sharing’ is made easier by the
use of communications buses.
It comes as no surprise then that the airbag
control system has (a) the ability to store data, and (b) uses a wide variety of
car sensors as part of its decision making. However, the use of the controller
as an Event Data Recorder (EDR) goes a step further – not only are fault codes
stored, but in some systems the input of a variety of car sensors is also
How did this extra step come about?
In the early 1970s the US National Transportation
Safety Board recommended that vehicle manufacturers gather information on
vehicle crashes using on-board collision sensing and recording devices. Since
1974 General Motors (GM) systems has recorded data for impacts that resulted in
the triggering of the airbag (a “deployment event”) while other systems were
also introduced that could additionally record “near deployment” events. In 1999
GM implemented the capability to record pre-crash data, that is, data is
recorded to a buffer on a continuous basis, the overwriting ceasing if a crash
occurs. Ford in the US started installing EDRs in one model in 1997 and by 1999
nearly all its US models were so equipped. A range of other manufacturers either
admit to some data recording or are looking to implement such strategies.
Rather than use airbag control systems to record
crash and pre-crash data, in US-manufactured heavy trucks the engine’s ECU is
used instead. On their diesel engines Cummins, Detroit Diesel and Caterpillar
all use electronic control systems which log driving data.
The precedent for recording data on a continuous
basis is well established in other areas of transport: in the US, regulations
requiring the use of Event Data Recorders are in place in aviation (from 1958),
marine (2000) and railway (1995) applications. The Transportation Safety Board
of Canada draws an interesting relationship between the presences of the EDRs in
different forms of transport and the number of fatalities occurring. “Motor
vehicle accidents on highways accounts for 93 per cent of all transportation
fatalities,” it points out, suggesting that it makes more sense to have EDRs in
cars than in any other form of transport.
The GM System
The information recorded by GM airbag systems
consist of data for both deployment and near deployment events.
A near deployment event (ie one where the airbag
doesn’t inflate) is defined as an event severe enough to ‘wake-up’ the algorithm
within the control unit. (An algorithm is used to analyse the severity of the
crash pulse, ie the control unit uses the shape and magnitude of the
deceleration pulse it is undergoing before deciding whether or not to fire the
airbag(s).) Two different systems are used by GM; one stores data on the near
deployment event which had the greatest change in road speed, and the other the
most recent near deployment event.
For both deployment and near-deployment events the
following are recorded:
Throttle Opening – the percentage that the
throttle is open, where 100 per cent is wide open. This information is sent by
the engine management system along with engine and vehicle speeds, so again is
recorded every second for 5 seconds prior to any event.
In addition, the number of ignition key cycles at
the time of the events and at the time of download is logged, as is whether the
passenger-side front airbag has been manually switched off.
One of the two GM EDR units is designed so that
150 milliseconds after the deployment algorithm has been enabled, all the data
stored in the memory is permanently written to EEPROM. It then cannot be erased,
cleared or altered – this type of device must be replaced after an airbag
(The Ford system records longitudinal and lateral
acceleration, deployment strategy of dual-stage airbag, seat belt use,
pretensioner operation and driver’s seat fore-aft position.)
1999 GM licensed the Vetronix Corporation to build a data retrieval tool for
their EDR. (Ford has more recently followed suit.) The Vetronix Crash Data
Retrieval (CDR) tool consists of both hardware and software. The system costs
GM airbag controller contains a full Event Data Recorder. The data logged just
before and during the crash can be read either directly from the module or if
the wiring in intact, from the car’s diagnostic port.
is a sample of the pre-crash data that is logged by the GM system, as read out
using the Vetronix Crash Data Retrieval tool. Throttle opening, engine and road
speed, and the on/off status of the brake switch are logged at 1-second
intervals for the 5 seconds before the crash.
the crash the change in speed is logged every 10 milliseconds, allowing a
detailed examination of the impact behaviour. The airbag system’s accelerometer
is used in this process.
GM data is logged in hexadecimal form and needs a dedicated reader to make sense
of the data.
As indicated above, heavy truck manufacturers are
implementing EDR in the engine management ECU. Detroit Diesel’s DDEC IV system,
for example, records vehicle speed, engine speed, throttle position and brake
and clutch switches for a period of time before and after a “hard braking event”
is detected. A hard braking event is user-adjustable but is normally determined
by a wheel deceleration of greater than 7 mph (~11 km/h) per second. The time
span and intervals between data collection vary with the different systems, but
data collection each second for the minute before the rapid deceleration and for
15 seconds afterwards is typical.
Caterpillar engines manufactured from 1996 are
equipped with EDRs and later models use an internal lithium back-up battery with
an expected service life of 4-6 years. A truck involved in a crash should
therefore retain the data for as long as six years after the event. The Detroit
Diesel system uses non-volatile memory which does not require a back-up battery
to retain its data.
Unlike car systems where - as one commentator
bemoans - “there are almost as many event data recorders as there are Original
Equipment Manufacturers and aftermarket suppliers”, all truck systems use the
same 6-pin Deutch connector and a common communication adaptor to work with the
The evidence appears strong that EDRs improve
crash analyses. The reconstruction process is not only simplified but the
accuracy of the reconstruction is improved, resulting in more detailed
conclusions. The so-called Haddon Matrix has been used to show the information
available with and without EDRs.
Information Available Without EDR
Calculate change in velocity
Environment after crash
Information Available With EDR
Conditions during crash
Measured change in velocity
Airbag inflation time
Automatic crash notification*
Automatic crash notification*
Automatic crash notification*
*Automatic crash notification refers to systems
with the capability of automatically alerting authorities (eg by mobile phone)
where and when an accident has occurred.
When you consider that a crash investigator
primarily has only vehicle damage and obvious physical signs like skidmarks
(less likely to be present with ABS) on which to make major judgements, it can
be seen that logged data on vehicle speed and other parameters can be enormously
So how good is the data collected via an EDR? The
answers to that are surprisingly broad; certainly there is plenty of information
available for someone who wants to fight the EDR evidence in a court of law.
However, on the other side of the fence, if used carefully, the data gained
post-crash from an EDR is of great use in helping determine the events that
occurred before and during the crash.
So what are some of the potential problems?
Vehicle speed, engine rpm, throttle opening and
brake status are logged only once per second – much too slow a sampling
frequency to be optimal when analysing many types of crashes. For example, did
the driver brake at 3.1 or 3.9 seconds before impact? The difference is major.
Additionally, these data are not synchronised with the start of the crash data,
so are potentially offset from crash data by up to one second.
The use of only five data points for each of the
parameters of speed, rpm, throttle opening and brake status can give a false
impression that the behaviour of these parameters can be validly shown by a
graph with these points connected by a straight line – but of course these data
might have been behaving quite differently between the discrete points.
Vehicle speed, engine rpm, throttle opening and
brake status are all dependent for their accuracy on car sensors and/or
switches. Vehicle speed sensor and throttle position sensors are often in error
- variations in accuracy of these parameters by up to 10 per cent is not at all
uncommon. This is a point that seems to have been overlooked by some researchers
in the field.
Much crash test work has gone into testing the
relationship between data gathered from EDRs and that gained through other
logging techniques. One approach is to measure the vehicle’s change of velocity
using the EDR and compare that figure with the crash test impact speed.
A series of Canadian tests showed that generally
there was fairly good agreement between the calculated and actual speeds – for
example, an actual impact speed of 40.3 km/h and a EDR-calculated speed of 42.4
km/h. Typically the EDR showed a slightly higher speed because it took into
account the bouncing back off the barrier that the car underwent after the
However, one test involving a 2000 Ford Taurus had
a significantly greater difference between actual (47.8 km/h) and EDR (53.6
km/h) speeds. The testers suggested that this discrepancy had been caused by a
spike in the acceleration/time curve caused by structural deformation in the
region where the EDR was mounted. A major discrepancy also occurred in another
test, this one where a 1988 Chevrolet Cavalier’s EDR lost power during the
crash. The independently measured test speed was 64.8 km/h but the EDR showed
In low speed tests another study found that the
shape of the crash pulse had a major affect on the accuracy with which the EDR
reported the change in speed that the car had undergone. Because it takes a
deceleration of over 2g before the algorithm is even enabled, a crash pulse with
a sinewave shape on the acceleration versus time graph gave different calculated
speeds to a crash pulse that looked more like a square wave. In other words, the
car is decelerating for a moment before the EDR starts to log – and so calculate
the speed change. This study found that the EDR underestimated nearly all the
collision speeds. The researchers reported that “since real vehicle-to-vehicle
collision pulses are probably shaped more like sine or haversine pulses than
square pulses, the speed change reported
[by an EDR]
likely defines a lower
limit for a vehicle’s speed change during a specific collision.”
However, away from the laboratory the usefulness
of the data – even with these reported inaccuracies – can be clearly
The 83-year-old male driver of a 2000 Buick
Century was negotiating a right-hand curve when he ran off the road, travelled
down an embankment into brush and tall grass, then crossed a level section of
lawn and a gravel driveway before colliding with two large rocks. The car came
to rest approximately 140 metres from where it had first left the road.
Time Before Algorithm Enable
Brake Switch Status
The pre-crash data obtained from the EDR (shown
above) indicated that the driver was operating neither the throttle nor the
brakes for at least 5 seconds prior to impact with the rocks. At the crash scene
the driver was lethargic; in hospital he failed to respond to treatment. An
autopsy showed that he had died from the results of a brain haemorrhage that had
occurred while he was driving – a diagnosis well supported by the EDR data.
Another example of the usefulness of EDR data is
in the crash of a 1998 Chevrolet Malibu, which ran under the back of a parked
truck. The car sustained severe damage to its bonnet and windscreen area,
resulting in a long crash pulse. The crash simulation software used by the
investigator estimated the change in vehicle speed to be 23 mph (37 km/h)
however through experience the investigator thought this figure to be low. A
reading of the EDR data showed a change in speed of approximately 50 mph (80
km/h), which to the investigator appeared much more reasonable.
The latter case shows that when other
investigative techniques might give speed indications in error by a very major
amount, discrepancies of a few km/h in actual versus EDR speeds are in many
cases of little importance.
While the potential societal benefits of the
universal fitting of EDRs are highlighted by many road safety researchers (see
the ‘Potential Benefits of Event Data Recorders’ breakout box below), many
drivers and some vehicle manufacturers are more concerned about the personal
privacy implications than the common good.
The US Federal Motor Carrier Safety Administration
has stated that it believes that the following standards should apply to
controlling access to EDR data:
Only the vehicle owner, or another party having
the owner’s permission, may access the EDR data. Exceptions would include
instances where a law enforcement official has a warrant in connection with a
At this stage none of those points has been
implemented, although truck owners can deactivate the EDR by means of setting
the deceleration threshold inappropriately, giving them some measure of control
over the data being collected. In a 2001 NHTSA report, Volkswagen is quoted as
stating that “due to high impact on privacy issues
owner’s choice at
[the time of the]
new car purchase or by dealer
Certainly there needs to be more public debate
about the privacy issues involved.
If the US success at causing Onboard Diagnostics
to be built into many of the world’s cars is repeated with EDR (and what
government could resist implementing the requirement for such a device,
especially when the financial cost to the consumer will be zero?), it’s very
likely that in 5-10 years time all cars will have accident crash logging.
So next time you’re involved in a car crash and
there is debate or uncertainty about the circumstances, think about the
implications of accessing the EDR – or at least making some serious enquires as
to whether one is fitted.
Benefits of Event Data Recorders
Real Time – Use of EDR data in conjunction with Automatic Collision
Notification systems would aid in quickly locating crashes and despatching
emergency personnel with better crash information in advance
Law Enforcement – Obtaining impartial EDR data from a collision would help
in more accurate determination of facts surrounding an incident
Government – Collection of EDR data facilitates government in further
regulatory initiative to help reduce fatalities, injuries and property loss
Vehicle Design – EDRs allow manufacturers to collect better real world data
to monitor system performance and improve vehicle design
Highway Design – The use of EDR data can assist in assessing highway
roadside safety and managing road systems
Insurance/Legal – Additional objective data provided by EDRs advance quicker
and fairer resolution of insurance and liability issues
Research – EDR data could provide objective databases of driver behaviour
and performance, as well as other research related topics
Owners/Drivers – EDRs can help fleet owners and drivers monitor vehicle and
driver performance to ensure the safe and efficient movement of people and
Multidisciplinary Road Safety Conference, 2001