USING NEXYAD ADAS MODULES
FOR AUTONOMOUS VEHICLE AND SAFETY/RISK ESTIMATION



USING NEXYAD ADAS MODULES FOR AUTONOMOUS VEHICLE AND SAFETY/RISK ESTIMATION

by NEXYAD


INTRODUCTION

The company NEXYAD developped software modules for Advanced Driver Assistance Systems :
. RoadNex (Road detection) : lane detection, detection of the borderlines of drivable area in the lane, detection of the surface of drivable area in the lane.
Sensor : camera (color)

. ObstaNex (Obstacles detection) : obstacles detection (if they have a vertical dimension or – inclusive – if they have their own movement)
Sensor : camera (N&B or color), accel, gyro

. VisiNex onboard (weather visibility measurement) : visibility measurement (quality and distance)
Sensor : camera

. SafetyNex : onboard road safety / risk estimation
Sensor : navigation map, gps, accel or car speed

Those modules were made to develop very efficient ADAS.
There are many ways of comining those modules, depending on the function that should be developped.

LANE KEEPING AND AUTOMATIC BRAKING : FOR CAR MANUFACTURERS AND TIER ONE COMPANIES

For this function, modules may be integrated in a rather complex way :
Nexyad Suite 1
Such an application needs to know where it works and where it doesn’t work (reliability). For that, VisiNex helps because it measures weather visibility and the nit is possible to know in which context artificial vision algorithms are efficient or not. It is also possible to switch setting parameters of artificial vision based algorithms using visibility characteristics, in order to expand the range of good performance of the global system (this is robustness).

NEXYAD applies a validation methodology called AGENDA (see papers in CESA Automotive 2014 in Paris and in SATETYWEEK 2015 in Aschaffenburg). This methodology is the onlt approach that allows to know what the system is supposed to do in a functional point of view, with measurable characterisctics of road scenes.
NEXYAD of course uses the NEXYAD ADAS validation data base : a part of this validation data base for artificial vision-based ADAS will be soon online for free (usable by every researcher or engineer in the world).

Note : the AGENDA methodology also provides a method to measure the similarity of a road scene in the validation data base anda current road scene : this is applied to estimate a confidence score.

SAFETY / RISK ESTIMATION FOR INSURANCE COMPANIES

SafetyNex measures the adequation of driving to road infrastructure characteristics.
It generates then a risk if the driver goes too fast when approaching a crossing road or a dangerous curve.
Of course, a poor visibility should lead the driver to drive slower.
In addition, there could be auxiliary inputs that would tell SafetyNex if there are obstacles on the pathway :
Nexyad Suite 2
This scheme is the same than the previous one but the outputs of RoadNex and ObstaNex are used INSIDE the scheme (they don’t provide an output of the global scheme).

DEMOS OF NEXYAD MODULES



REFERENCES

Validation of Advanced Driving Assistance Systems by Gérard Yahiaoui & Nicolas Du Lac
CESA Paper by Gérard Yahiaoui & Pierre Da Silva Dias
Road detection for ADAS and autonomous vehicle
Using the NEXYAD road detection (RoadNex) to make obstacles detection more robust
Real Time Onboard Risk Estimation Correlated with Road Accident
Visibility Measurement for ADAS and Autonomous Vehicle