Hello,

my name is Aubanel Monnier, I am the director

at Ansys customer excellence for systems in

Asia Pacific and

today I'm going to talk to you about our

solutions

for

NCAP testing.

Every year,

more than 1 million people are

killed on the roads in the world. A way to

bring

that number towards zero facilities is to

improve the safety

of the vehicles.

This is what NCAP is contributing

to by rating

the safety of the vehicles by giving them

stars between

zero to five stars.

And for example if you compare a vehicle that

has

five stars to another vehicle that has two

stars.

The five star vehicle lowers the risk of

fatal injury

by about 68%,

and if you compare a 5

star with a four star rating vehicle,

it's 12%.

NCAP is composed of nine regional bodies

who are conducting independent research and

testing programs.

They compare the safety characteristics of

different motor vehicles and

rate from zero stars to five stars.

5 star vehicles are available in all the regions

,

typically more than 50% of the tested

vehicles reached the

five-star grade.

For example, in Europe it's 72%

in two in 2017.

But there are still cars that

receives zero stars and that makes NCAP

a marketing and

sales argument for car manufacturers.

Here are the nine regions into which NCAP

has developed

direct presence, Australia,

Southeast Asia,

China,

Europe.

You will see Japan,

Korea.

Latin America, Caribbean's, and India,

and for which local test tracks have been

developed to

conduct the different traffic scenarios.

Fünf, Vier, Drei, Zwei, Eins.

This video is an advertisement broadcasted by

Mercedes Benz.

It shows the work that they did over 60 years

to improve the passive passive safety of their

car.

We can see the evolution of the methods used

to

test vehicle.

Improvement in sensing and measuring

capabilities.

Multiplication of testing scenarios.

Introduction of digital crash tests.

Operation of new passive safety devices such

as airbags.

And ultimately use that as a strong sales

argument.

Passive safety has been largely studied in

the past years,

Ansys provides complete workflows,

including for example,

Ansys LS-DYNA,

which is the world leader in crash simulation

.

But cars are becoming smarter and are

taking an

active play to increase the safety of the

real users.

It starts in 2002 with things as simple as

seat

belt reminders,

and then smarter systems appear like

electronic stability control and

speed assistant systems.

In the last years,

first ADAS systems like lane keeping and

automatic emergency braking.

Of course, it's only the beginning.

It's important to keep in mind that

governmental regulations and

the NCAP ratings are two different things.

They don't influence each other.

Compliance to local regulation is mandatory

to put the car

on the road. NCAP ratings,

on the other hand, is meant to be a marketing

differentiator.

Scenarios and ratings are different in the

different regions.

As an example,

the USA mandates a crash worthiness rating

label on cars for sales.

On the other hand,

An NCAP rating is not mandatory in

Europe,

but its absence is viewed as suspect by buyers

,

explaining why 80% of new models are sent for

a

Euro NCAP rating by the manufacturers

themselves.

A good safety rating is good for your business

.

NCAP of course,

following this trend of an increasing

role of the active safety systems, at the

beginning of

this year,

they published their five years road map

detailing the testing,

additions and improvements they are planning.

So what are your challenges?

Let's have a look at the road map NCAP

publishes.

We see a number of automatic emergency

breaking systems and

steering systems.

The development of the protocols started a

couple of years

ago and the protocols were released last year

.

And the implementation is starting now.

Such assisting or automated system will

have to be tested

for a wide variety of use cases and scenarios

and

that means that your systems and cars will

have to

demonstrate their usefulness and capabilities

in an always increasing number of

situations like

approaching a stationary car,

approaching a slower moving car,

approaching a car that is breaking,

having a car cutting in your lane, or a car

cutting out and unveiling another slow car in

front of

you.

Steering to avoid obstacles on the road,

steering to follow the road curve.

So how can we help you?

What we do propose is to let you test and

rate EuroNCAP virtually.

And improve the robustness of your ADAS

function whether it's

emergency breaking, pedestrian crossing, closed

cut in etc.

The aim is to test the functions algorithm

and to

be able to evaluate the vehicle dynamics

reaction in context.

In order to do so,

we need to perform closed loop simulation

with the function

of the task,

the vehicle dynamics and the driving scenario

.

We can approach the NCAP certification in

three steps.

First,

support the autonomous driving function

developers with a fast development

verification and validation loop to

execute many tasks to tune their control

algorithms and ensure

their robustness.

We have augmented our Ansys VREXPERIENCE

driving simulator with NCAP

development scenario package containing a

digital representation of the

NCAP testing environment and scenarios.

In the second step, the V&V engineers can use a

certified version of this package to

guarantee faithfulness of the

digital testing.

And finally,

we can put the car on the proving ground

with a very high level of confidence in the

correct

behavior of the function and test.

The NCAP development scenario package will

help you in

deriving any NCAP scenario from the NCAP

specs. Of course you won't have to start

from

scratch as the package

come with the library of readymade scenarios

.

You'll be able to make quick changes and try

different

parameters on the fly to understand their

influence.

As the number of parameters to play with is

quite

significant,

you'll be able to use HPC to run massive

scenarios

and you'll be able to simulate a variety of

environments

such as various lighting conditions.

So. Let's have a look at what the NCAP

developments value package provides.

A set of test protocols for front collision

warning and

automatic emergency breaking functions

covering a variety of use cases.

Car to car,

adult or child pedestrians, and bicycles.

It has a high fidelity to the NCAP

environment,

as the 3D modeling follows the NCAP

specifications and

the material optical properties are measured

with specific optical measurement

devices,

enabling physics-based simulations including

their sensors.

From the official NCAP protocol we have

developed 14

scenario families,

each one containing more than 10 variables

and at least

10 key performance indicators.

These permits to build about 150 NCAP test

cases

and to produce and execute millions of tests

for control

algorithm tuning and robustness test.

To test a ADAS function,

we have to connect it both to its simulator

sets

or efforts and it's activated outputs.

In this first video,

the sensor detecting the pedestrian presence

is not connected to the AEB function at the

test.

Obviously the system does not see the

pedestrian and runs

over him.

If we close the loop by connecting the sensor

to

the function under test,

we can witness the correct execution of the

function.

The car stops before hitting the pedestrian.

Now let's introduce variability in our

simulations.

Let's just choose one parameter.

In this example,

it's the initial speed of the overtaking

car,

and we're going to set a range from 50 to

80 kilometers per hour and an incremental

step of five

kilometers per hour here.

And so this is going to generate 7 scenarios.

We can execute them one after the other or in

parallel batches with the adequate setup,

and for each of these cases we will be able

to collect key performance indicators to

understand the cause effect

relationship between the parameters and the

KPIs.

This works of course for a wider number of

parameters

and can quickly generate a very

a very large number of scenarios.

High performance computing setups can help

executing these large number

of cases.

Smart sensitivity analysis tools such as

optiSLang can also

be connected to reduce the number of actually

executed tests

while providing a very precise understanding

of the effects of

the parameters over the KPIs.

The NCAP digital certification package

certified by UATC augments

the previous package with the true

reproduction of actual tests

on the proving ground.

There is less than one-centimeter error in

target trajectory

reproduction.

This package is really dedicated for a

digital pre validation

and produces a prefilled report to present

the vehicle to

the NCAP certification.

Once the ADAS function is validated,

the limitations of sensing capabilities of

the systems need to

be taken into account.

To do this,

we're going to replace in our simulations

ground truth sensors by physics based sensors.

We can easily start with NCAP regulation

inspired geometries.

And extend the materials with precise optical

parameters available in

the visible range and near infrared and

including retro reflective

material.

With this we can evaluate very precisely the

sensor output

in static simulations thanks to Ansys

capabilities.

We can also,

at the expense of a slightly reduced

fidelity,

simulate the sensors in a real time dynamic

simulation,

and for a wide variety of sensor types.

For these simulations,

we do provide all the needed assets.

Pedestrians with and without safety vests

most frequent standardized scenes

in the automotive industry, a variety,

of road intersections,

and typical models for headlamp beams.

We also provide a library of useful optical

materials.

Road asphalt and markings,

which have been adapted from French

academic publications.

Road signs with the standard industry coatings,

police plate definitions that have been

measured from real plates.

Safety vests featuring standardized

retroreflective and fluorescent materials,

and we complete all that with

typical car paintings and environment and

elements.

Inserting the sensor into the simulation loop

permits to understand

the effect of its inherent limitations.

Let's have a look at an example.

In this very simple scene,

we're going to have pedestrians crossing

while the car is

cruising at increasing speeds.

On the left hand side we can see the output

of the camera sensor.

On the right hand side is the output of a

perception algorithm executed over the

simulated sensor output.

As the visibility is very clear,

we can see that the car recognizes the

pedestrian from

far enough

to have the time to stop.

While we analyze the results,

we see that only at the highest peak the

AEB is not capable to stop before hitting the

pedestrian.

Now,

if we change our conditions and rerun the

same set of tests but at night time.

We have street lamps and low beams lighting

the scene.

The perception will be able to detect the

pedestrian much

later and hence collisions will happen at

much lower speeds.

This capability to work at multiple levels of

fidelity is

something quite important.

You do 80% of the testing on the pure

ADAS control software using ground truth or

stochastic sensor models.

Once this is validated,

you're going to do the 20%

of remaining testing to understand sensor

limitation effects.

Note that the simulated sensor limitations

have to be correlated

with the safety of the intended function

analysis.

As a rule, the numbers of sensor is bound to

significantly

increase as

levels of autonomy are also raising.

We expect the role of physics based

simulation of sensors

to become more and more important.

So let's summarize with your benefits.

As of 2020 NCAP testing for two functions

will

cost you 4 new vehicles and about six

millions of

US dollars.

In 2025,

a typical car will rather have about 12

functions to test.

You'll still need 4 physical cars,

but the higher number of functions will bring

your cost

in the 20 millions range.

That's a three times increase of cost in less

than

five years.

Without Ansys, you can reduce by 50%

your physical tests,

reduce the risk and secure you five

stars rating.

If we illustrate by a concrete example here,

the process used by Renault Nissan.

Up to 2018 they were developing a first

prototype that

they were using to test the function

internally over the

314 EuroNCAP use cases.

Then it would develop a second prototype and

validate it

the first time on the EuroNCAP proving

grounds.

With this test results and additional testing

,

they would fine tune the system until it is

ready

for an official EuroNCAP second testing.

Since this year they had diminished that

physical testing cost

and time by 50%

by replacing the first NCAP test and the

tuning

phases by simulation and digital optimization.

They can get their cars on the market faster

and

at a lower cost.

Well,

we reached the end of this presentation.

I hope you found it interesting.

Don't hesitate to contact us if you want to

know

more.

Thank you for attention bye-bye.