NEXYAD Automotive & Transportation Newsletter #26, March 24th, 2019
New Important Reward for SafetyNex
– SafetyNex won Road Safety Innovation Prize
– Road Detection RoadNex New Update
– Go from Level 3 to Level 5 with SafetyNex
– High Level Fonctional Specification of a Self-Driving Vehicle (Point to Point, any Destination): What about Driving Risk ?
– News about NEXYAD in Global Fleet
– Nexyad CEO Speaker at two conferences on AI
Nexyad presented SafetyNex, Artificial Intelligence based software component that computes at each moment the Driving Risk (like the teacher besides you the day of the driving licence) that the driver is currently taking :
. for human drivers, a vocal alert makes you aware that your taking too much risk and then driver can slow down and avoid emergency situations
. for Autonomous Driving systems, it is possible to servo-control Driving Risk and keep it under a maximum accepted risk value : AD system becomes adaptive to risk.
All Nexyad Team is proud with this Innovation Prize !
Some companies try to detect freespace with a regular camera, without IR (SWIR etc …), without fusion with LIDAR, just because it allows low cost applications especially for aftermarket. Here is an example of the NEXYAD module RoadNex. One can see that there are a lot of cracks, patches (highlights and dark zones), and that RoadNex still detects free space in front of the car. Especially useful in the roundabout (see video).
Note : It runs on a regular chip (real time on a regular smartphone, for instance). It means that cost of deployment is “cost of software only”, NO NEED for a special smart hardware architecture (NVIDIA …) or a special camera. This is the differentiation of NEXYAD : we do not try to provide “perfect and beautiful” detection (that is already achieved by MOBILEYE for instance, but we think, beyond market price for markets we target), but rather an efficient quick detection for some valuable functions at market price. Our free space detection is currently under integration into dashcams.
Yes, level of autonomy is not a matter of number of use cases taken into account during fine tuning or deep Learning, and during validation. Indeed, a feedforward chain cannot be trusted anyway because it is not adaptive. The number of use cases is quite infinite and there will alsways be a “hole” in your scenario crossings
So once your feedforward AD system chain seems to work at 94% (that’s what Elon MUSK said about new TESLA), it is not necessary to go on developing it. Yes, it is time to make it adaptive by integrating this AD system into a loop of safety monitoring. Of course, everyone thinks of functional safety (covariance, etc), and it must be done, but again it is not enough. We propose to measure at each moment the Driving risk that the AD system is taking (road safety) in order to adapt decision automatically to keep risk under a maximum accepted value. This is easy to do using our software component SafetyNex as an added layer to your existing AD system.
See figure below.
(Point to Point, any Destination): What about Driving Risk ?
A general usage self-Driving vehicle is an object that picks you at a point A (that may be anywhere) and takes you to a point B (that may be any reachable point from A), as fast as possible, in comfort and safety. And that’s it (main features).
a . as far as possible : it is the actual efficiency of transportation, as mobility is a key element of Economy
b . comfort : it is the fact that the travel is enjoyable
c . safety of Driving : risk of road accident is few because the vehicle controls the Driving risk that it takes. Note : we consider that the object in itself is already safe (functional safety solved).
How to control those 3 items ?
a . chronometer, speed measurement
b . score of comfort : noise, vibration, user friendly HMI that cancels stress (infotainment), entertainment, …
c . score of Driving risk (in terms of ROAD safety) that the vehicle is taking while Driving : it already exists and it is the module SafetyNex, developed by NEXYAD (validation on 50 million km).
This gives a clue about how SafetyNex is used by Tier One Companies for their AD system.
Nexyad uses artificial intelligence to improve road safety
“75% of road accidents happen because of driver behaviour that is not adapted to local infrastructure,” said Gérard Yahiaoui, CEO, Nexyad. His company sells software components that use artificial intelligence to improve road safety. Fleet Europe wanted to know more about where his company came from and where it is going.
In June 2013, Nexus Communication has launched GLOBAL FLEET, the first networking and crossmedia platform for multinationals willing to optimize their fleet management through globalization. It provides professionals with in-depth knowledge and up-to-date information on global and regional fleet management as well as on BRICs and other economies on the rise.
– at round table on Artificial Intelligence & Industry Transformation for Startup Innovation Challenge March 27th 2019 (Paris).
See more : https://sommetdesstartup.challenges.fr/
– at ROTARY Digital Meeting april 10th 2019 (Paris).
See more : http://www.rotary-paris-ouest.fr/offres/gestion/events_146_45209_non/rotary-digital-meeting.html