Insurance & Technology is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

News & Commentary

10:45 AM
Wayne Gilbert, Risk Technology Ltd.
Wayne Gilbert, Risk Technology Ltd.

Solving Problems with Crash Detection Technologies, Part 2

The key objective has been to test the calibration, consistency and rejection of false positives, for incidents occurring at low speeds.

Editor's note: This is the second part of a two-part series. Last time, Gilbert outlined the issues with existing crash detection technologies. In this installment, he outlines how new developments work, how they've been tested and applied in the real world, and how it works in concert with other data. For the first installment, click here.

Today's crash detection technology uses two inputs to detect and confirm an impact. The normal G-force measurement, provided by accelerometers, is qualified by sensing the acoustic waves created at the time of the impact. Only when both events occur simultaneously at pre-defined levels, is a crash event reported.

Few drivers will be unaware of the noise created when the plastic materials at the front or rear of a car are hit by a supermarket trolley! The familiar sound is also transmitted through the structure of the vehicle and detected by a sensor. In this example, the energy transmitted is low and very little G-force will be seen by the detection system. Conclusion: no crash (but possibly some scraped paint!).

A hard impact, created by a larger body such as another vehicle does far more than generate acoustic waves. The impact is transferred through the rigid structures of the vehicle and creates acceleration forces that can be measured.

This condition is true even when a small acceleration for is generated, as in the example. In this case, a low speed crash generates both a measurable g force combined with an acoustic signature. This means that a detection threshold can be set that is lower than the limit that could be used to identify an event only on g force.

The trigger thresholds for both G-forces and acoustic input are configured by software and may be different for larger and smaller vehicles and may be actively learned by the devices. These include measures of both the strength of the impact and its duration.

A similar process is calibration and sampling process is undertaken by both acoustic and accelerometer arrays. On installation, these sensors learn the orientation of the unit (front, back etc.) and align with the x, y and z axes of the vehicles automatically. Critically, acoustic sensing detects structure borne waves, not sound from the impact (although the two events do occur simultaneously). Therefore loud music or the kids fighting on the rear seat will not trigger the sensing system.


Device testing and system validation has focused on testing real-world conditions of discrimination. The key objective has been to test the calibration, consistency and rejection of false positives, for incidents occurring at low speeds.

Extensive testing of the technology has already taken place, using calibrated sensors and a controlled environment (Sled Tests). Pre-production devices fitted with the technology have also been tested extensively on road trials. The program took a number of devices and fitted them into test vehicles in a controlled environment. Test vehicles were then be exposed to a range of test collisions at low speeds (simulating the worst-case situation) to evaluate the behavior of the devices and to ensure that detection of the events occurs.

Test vehicles were also fitted with laboratory instruments and cameras to record all the tests in detail. The data output from the devices was sent "over the air" to servers to ensure end-to-end functionality. Vehicles were driven over a range of prepared test surfaces to ensure that noise transmitted through the vehicles suspension did not cause false positive to be reported.

Results and Applications

'Real-world' test conditions, to prove the systems discrimination of events, showed excellent consistency and rejection of false positives.

It was necessary for the technical solutions to be simple, effective and critically low-cost. Extensive research went into the selection of appropriate devices and the setting of parameters to best effect. The results showed that the technology was able to deliver a solution that has the highest sensitivity and best crash discrimination, validated by a world authority.

Today, work is being carried out to use data extracted from the vehicle itself. For example, were the brakes applied at the time of the incident? What was the throttle position before the event? Were the seat belts fastened? As a next step, the technology will include data about the road conditions; weather; traffic density and the localities record for incidents will be included. Such information will help insurers and car investigators not only understand the actions of the driver but also understand any environmental factors which will increase the accuracy of the data and avoid possible misinterpretation.

About the author: Wayne Gilbert is CTO of Risk Technology Ltd, a telematics product provider based in the UK.

Register for Insurance & Technology Newsletters