Go to Global presentation of applications for motor INSURANCE
Driving risk is the risk taken by the driver at every moment during driving task.
Driver maybe a human being or the artificial intelligence (autonomous driving)
RELATIONSHIP BETWEEN DRIVING RISK AND ACCIDENT : THE “S” CURVE THEORY
Let’s say in a manufacture there is a very dangerous machine that may grind up your hand. If you are 10 km away from the machine, risk is “very” low. If you are 1 km away from the machine, risk is the same. If you are 10 m away from the machine … risk is still very low … but if you come closer (let’s say 10 cm), suddenly risk becomes high ! This is not linear. In road safety, the Artificial Intelligence algorithm SafetyNex estimates 20 times per second the driving risk you take, and many people ask about relationship between “risk you take” and “accident”. This relationship is not deterministic (probabilities must be used) : risk is not linked directly to accident but rather to accident frequency (or probability) … and the relationship is a non linear curve called a “S” curve as shown on the figure below. It is possible then to use it to alert human driver (Vocal Driving Assistants) or to control autonomous driving (Autonomous Vehicle) in order to keep risk under the threshold of the “S” curve or not too far after the threshold. SafetyNex was calibrated in order to have 95% of accident frequency just after the threshold (validated on 50 million km).
A regular good driver will generate a SafetyNex driving risk(t), 20 times per second, between 0% and 70% with a maximum likehood between 30% and 60% depending on driver’s self confidence.
SafetyNex alerts at 90% and takes into account mostly [80%-100%] of driving risk slot to compute Safe Driving Score at the end of a trip, according to the S curve theory.
Contrary to what many companies do (risk scores based on statistics and data analysis), SafetyNex by Nexyad uses the true scientific method proven and validated : theory of risk developed by Frank E. Bird : “the triangle of risk”
This approach is in use in many domains confronted to risk management as the firemen, the FBI, in Nuclear Plants, etc.
NEXYAD is the first and only team that applied it to road safety.
It is very obvious, the same driving behaviour on a disused airport, on a highway, in front of a school, in a dense city approaching an intersection, etc… do not correspond to the same driving risk ! And no one can deny it.
If you measure driving behaviour without knowing driving context, it is impossible to compute driving risk. Training drivers not to do severe braking is very dangerous: they will hesitate even if a child surrounds in front of the car. Severe Braking is sometimes absolutely necessary. And that is why car manufacturers develop more efficient braking systems at each car generation (drum brake, disc brake, ABS, EABS emergency assisted braking systems), spending a lot of money for it !
If you take safe driving coaching course, they will teach to break your speed very harshly in case of any doubt. Road Safety should not use “intuition” of non-expert people or “belief”, but science and facts.
Just to make sure that our explanation makes sense, let us compare on a very simple case of urban risky situation (STOP sign) the harsh braking detection and the SafetyNex alert.
– Use case 1 : the driver arrives too fast approaching a stop sign. Instead of anticipating, this driver sees the stop sign at the very last minute and brakes very harshly. It is possible to put a threshold on acceleration value and then detect the harsh braking. This harsh braking detection obviously corresponds to a driving risk, and the figure 1 shows when it happens on the pathway.
SafetyNex predicts the next locations of the pathway, and estimates in real time if driving behaviour is adapted or not to the complexity and singulatities of road infrastructure ahead.
Because SafetyNex cannot predict that the driver is going to brake (the behaviour does not show it), SafetyNex driving risk rises as shown on figure 2.
One can notice that as soon as SafetyNex understands that the driver finally brakes enough to stop the car, then driving risk falls down very quicly. Driving risk rises BEFORE potential accident : first you take risk, and second you may have an accident.
SafetyNex and harsh braking criteria both detect that the driver took risk in such a situation : if you count alerts, you have 1 for harsh braking and 1 for SafetyNex.
What is the difference ? Well SafetyNex anticipates and then SafetyNex can alert the driver a few seconds before stop sign, letting time to slow down, where harsh braking criteria only monitors but cannot be used to alert because it has no anticipation feature.
– Use case 2 : It is exactly the same situation, but the driver doesn’t brake at all
Because there is no braking at all, the detection of harsh braking is not possible, as shown on figure 3 :
This driver that is much more dangerous than the previous one is not detected by harsh braking detection. Harsh braking detection is not effective on this use case.
SafetyNex works exactly the same way than in the previous use case : because it is not possible to forecast that the driver will brake, then risk rises as shown on Figure 4 :
Risk falls down only when situation is left behind : after crossing the stop sign, it is too late for risk assessment, it is an « alea jacta est » situation, you will have an accident or not depending on luck, there no possible anticipation.
. Conclusion :
Everytime that harsh braking corresponds to a real risk, SafetyNex also gives an alert. And SafetyNex still gives alerts in very dangerous situations where harsh braking detection is not effective.
SafetyNex is simply the next generation for driving risk assessment, you can now forget harsh braking detection and leave it all behind.
One more point, SafetyNex risk rises BEFORE dangerous situation, then you can alert driver and ACT on accident rate reduction.
SafetyNex can reduce accident rate by 20% for a regular inattentive driver (your ROI evaluation is easy to do).
Trick of numbers : stats have a meaning ONLY on very large numbers.
A black spot is an area where 5 falatities occured the 5 last years: black spot change every year.
Alerting on black spots: USE CASE
Road #N13 (France) between Saint Germain en Laye (NEXYAD office) and Paris La Défense : 160,000 vehicles per day. Then 58,4 million vehicles per year. Then 292 million vehicles in the last 5 years.
Let’s say that there is a black spot on this segment: a maximum of 5 vehicles (maybe less) had an accident with fatality… because one accident can lead to several fatalities.
Vehicles with no fatal accident : (292 million – 5) vehicles that didn’t have this problem…
The black spot is mostly a safe zone.
Observed frequency of fatal accident on the black spot: 5/292 million = 0,000000017 which means that on a black spot, fatality almost never happens.
Do you really want to be alerted for something that almost never happens?
=> Black spots recording (on maps) is not relevant for road safety alerts!
Modulation of Driving Risk with speed profile approaching STOP signs
Note : Driving Risk rises before STOP signs, and depending on speed (anticipation)
STOP sign is only an example, of course SafetyNex works the same way for curves, winding roads, intersections, school zones, pedestrian crossings, etc …
= Red alert (and for an active automated braking car, it would be trigger to anticipation braking)
Here is a demo of SafetyNex alerts during a real trip (using a simple implementation of SafetyNex API into a smartphone App made for demo ).
This App alerts driver BEFORE a dangerous situation, letting time to slow down and avoid a potential emergency situation (application to Vocal Driving Assistant for Automotive industry, or to Accident Prevention for Insurance / Fleet risk management). If you work on robotized car (ADAS or Autonomous Driving, just fancy that SafetyNex alerts your AI system intead of the human driver).
BEWARE with the statistics : “94% of severe personal damage accidents are due to human errors” doesn’t mean that you’ll save 94% of severe accident with autonomous driving : drivers do not only make mistakes they also drive well (1 accident every 70 000 km to 100 000 km, 3 dead every billion km – OCDE) … It is important to study also good driving and near misses (when driver has the right behaviour to avoid accident or to mitigate severity)… That’s what NEXYAD did during 15 years of research programs on road safety ^^ (that led to SafetyNex). See image (if you do not provide the “green” features, you will lose lives more than you gain with your driverless car. Our AI algorithm SafetyNex was made for this.
And SafetyNex is a XAI (eXplanable artificial Intelligence) so you can trust it (see Darpa research program on XAI).
Nexyad made 5 anime videos of 1mn each:
