SafetyNex driving risk profiles may solve the problem of hypovigilance and fatigue detection
SafetyNex is a nomadic real-time risk estimation system, based on Artificial Intelligence (AI). The system has been described in detail in previous publications and uses the key concept of "near-accident" or "quasi-accident", and is a result of 15 years of collaborative research with road safety experts and researchers. See http://www.safetynex.net
The main competitive advantage of SafetyNex is that it allows, since the risk is estimated in real time, to warn the driver (vocal alert), and thus to allow driver to avoid accident. Studies show that SafetyNex can reduce accident rate by 20%, which represents for insurers and fleet managers a consequent increase in margin.
But of course, SafetyNex also records usage and risk profiles. These profiles provide an the behavior of the driver, or more
precisely, his/her ability to regulate driving task consistently with danger. No need to record large volumes of data
(accelerations, etc ...) which in reality are not data (these are signals) for a possible back-office analysis, SafetyNex
provides exactly the interesting data.
One can see that the risk profile is very interesting : driver regulates around a given risk in a quasi gaussian way. Alerts are high risk slots and they happen when driver did not anticipate well.
Psychologists and human factor researchers model the task of driving with the concept of "mental load". The mental capacity of
driver can be seen as an « empty glass ». Each task fills this glass partially with a quantity called mental load .
When the number of tasks to be performed increases, the overall mental load exceeds the capacity of the glass. This type of model is
used to study the relevance of new driver assistance systems that involve new tasks of interaction between the vehicle and the driver..
. Assumptions :
We wish to formulate a corpus of assumptions, and encourage researchers in human factor to study them in order to confirm,
or not the following sentences:
. The risk profile is an image of the control task of driving : the driver must adapt his/her commands to follow the shape of infrastructure and regulate the risk. The risk profile is therefore an image
of the driving style in the sense that it is the fingerprint of the mode of regulation used by the driver to adapt driving style to infrastructure.
. The maximum likelihood of the risk profile is an estimator of the risk taken voluntarily (or consciously by the driver).
We assume that for a given driver, the mental load required for driving increases with this maximum likelihood.
As a result, novice drivers who naturally use a big mental load to drive (as they have not yet acquired all automations) will tend to "slow down" and thus position their maximum likelihood at low risk.
We also hypothesize that the experienced driver, with efficient automation that unloads his work
Experienced driver can roll a little faster without being overwhelmed by the amount of information to be processed, which is characterized with a maximum of likelihood of the higher risk profile.
This would mean that the driver would regulate his voluntary risk so that his mental load would be "sufficiently high" so as not to "forget that he/she is driving", and "sufficiently low" not to be
overwhelmed or tired by the task of driving.
If this is true, then, during the driving task, a shift to the left of the maximum likelihood should be seen as fatigue increases. And if the infrastructure has difficulties, then high risk alerts should also
On the contrary, we believe that temporary hypovigilance (typing a SMS, searching for something in the car, ...) should lead to high risk alerts without really changing the overall shape of the risk profile.
If those assumptions are at least partially validated, this means that SafetyNex, in addition to being used as a real time driving risk assessment system, could also be used to detect driving fatigue and hypovigilance.