Real time driving risk assessment for autonomous driving systems improvement
Autonomous driving systems usually have the following structure:
This chain takes a road scene and delivers an action: for 1 road scene, 1 action.
Then, if the road scene is different from those imagined by engineers, the action is potentially wrong and it is dangerous. This chain cannot adapt to new unknown situations.
NEXYAD has developed a unique innovative Global Road Safety platform that computes in real tim at each moment (20 times per second), the driving risk that vehicle is taking in its environment.
This driving risk varies from 0 to 100 and calibration was made to represent more then 95% beyond 90.
First application of this NEXYAD technology is to alert driver when risk riss a little too much, and it has been validated that vocal alerts on risk threshold can reduce accident rate by 20%, saving more than 200€ per year per vehicle and saving lives.
The idea for Autonomous Driving systems is to make sure that this AD system is "aware" of the current driving risk. Being aware of the driving risk, AD system can then ADAPT its action in order to keep risk under a max acceptable value.
Then, for the first time, an autonomous vehicle can become adaptive to completely new unknown situations. This is a good lead to go from autonomy level 3 to level 5.
New structure is then:
Autonomous car manufacturers (i.e. TESLA, VOLVO, ...) can now use this technology and greatly improve their vehicles without any complexification of their level 3 AD system.