ADAS/AD Development, Validation and Homologation Public 1 Tobias - - PowerPoint PPT Presentation

▶
adas ad
SMART_READER_LITE
LIVE PREVIEW

ADAS/AD Development, Validation and Homologation Public 1 Tobias - - PowerPoint PPT Presentation

Nov 26, 2022 304 likes •647 views

Public relations Public & politic reactions in case of accidents Customer acceptance Technical Cyber security Functional safety & Reliability in all Ethic conditions (Traffic, Weather & Road) Infinite

slide-1
SLIDE 1

1 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

Technical

  • Functional safety & Reliability in all

conditions (Traffic, Weather & Road)

  • Infinite high number of test cases
  • Complex and new technology
  • Time & Cost pressure during development

Public relations

  • Public & politic reactions in

case of accidents

  • Customer acceptance
  • Cyber security
  • Ethic

Legislative

  • Homologation criteria
  • Responsibility in case of accident

The rocky road to

ADAS/AD

Development, Validation and Homologation

slide-2
SLIDE 2

2 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

Level 5 Future Now

Responsibility

Driver’s License Regulations

? ?

slide-3
SLIDE 3

3 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

“The critical path to introduce autonomous driving vehicles will not be the technology but the development of a metric which empowers for an approval”

  • Prof. Dr. rer. nat. H. Winner, November 2013

A lot of progress concerning the

TECHNOLOGY

is visible, but:

… and the next question will be how to execute scenarios, test cases etc. of this metric in a limited time

slide-4
SLIDE 4

4 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

“The biggest hurdle is validation to confirm that the system does not cause failures. One has to execute 250 million test kilometers” Bosch Executive Director. Same approaches as 36 years ago?

Mercedes-Benz 190 E (1982) - Demonstration and advantages of ABS

New and Innovative Approaches in

Testing and Validation

are necessary!

slide-5
SLIDE 5

5 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

Facing the Challenges of Development and Homologation of Autonomous Driving Using Virtual Approaches

Driving Software and Digital Driver Legislation, Regulations and Conformity Vehicle Validation and Self Driving Vehicle

§

slide-6
SLIDE 6

6 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

Regulation for Homologation and Approval of AV

  • Dr. Houssem Abdellatif

Global Head Autonomous Driving and ADAS

slide-7
SLIDE 7

7 2019-03-20 TÜV SÜD MOBILITY | PERSPECTIVES FOR HAD HOMOLOGATION

HOMOLOGATION Qu'est-ce que c’est?

slide-8
SLIDE 8

8 2019-03-20 TÜV SÜD MOBILITY | PERSPECTIVES FOR HAD HOMOLOGATION

Maneuver / scenario data base

  • description
  • parameters
  • fail/pass criteria (KPI)

subset for homologation subset for physical testing subset for simulation testing

real vehicle virtual vehicle

results: m vehicle variants X n scenario variants X …

test track simulation

Take a SIMPLE IDEA … A 6 Points Approach to realize this Idea to empower for Approval

slide-9
SLIDE 9

9 2019-03-20 TÜV SÜD MOBILITY | PERSPECTIVES FOR HAD HOMOLOGATION

slide-10
SLIDE 10

10 2019-03-20 TÜV SÜD MOBILITY | PERSPECTIVES FOR HAD HOMOLOGATION

slide-11
SLIDE 11

11 2019-03-20 TÜV SÜD MOBILITY | PERSPECTIVES FOR HAD HOMOLOGATION

slide-12
SLIDE 12

12 2019-03-20 TÜV SÜD MOBILITY | PERSPECTIVES FOR HAD HOMOLOGATION

slide-13
SLIDE 13

13 2019-03-20 TÜV SÜD MOBILITY | PERSPECTIVES FOR HAD HOMOLOGATION

Change This!

slide-14
SLIDE 14

14 19-03-20 TÜV SÜD MOBILITY | PERSPECTIVES FOR HAD HOMOLOGATION 14 2019-03-20

slide-15
SLIDE 15

15 2019-03-20 TÜV SÜD Auto Service GmbH | GTC Europe 2018

Scenario data base

  • description
  • parameters
  • fail/pass criteria (KPI)

subset for approval

subset for physical testing subset for simulation testing

real vehicle virtual vehicle

results: m vehicle variants X n scenario variants X …

test track simulation test bench

subset for testbench testing

Combining Tools (Tool Chain) for Concise Approval

slide-16
SLIDE 16

16 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

Virtual Testing & The Toolchain

  • Dr. Tobias Dueser

Department Manager Advanced Solution Lab

slide-17
SLIDE 17

17 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

A seamless but open toolchain with new approaches is necessary

Simulation Same Results Virtual TESTING Proving Ground Real World

Maneuver, Scenario, Test Case Abstraction incl. KPIs Environment Abstraction Vehicle Abstraction

The most efficient validation will be done by those who will use the best combination…

slide-18
SLIDE 18

18 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

Proving Ground and Real World Finalize and Confirm Virtual TESTING: Integrate, Analyze and Improve Simulation: Scale Testing of Variants

(Vehicle Configurations and Scenarios)

Insights: Approach for the best combination…

reproduce analyze Optimize/ Improve

Remark: Focus of this slide is system (vehicle) validation. In addition there will be component test beds like HiL, Sensor Test Beds, etc. + specific test cases which cannot be done in simulation + specific test cases which cannot be done in simulation and/or Virtual Testing

Simulation Virtual TESTING Proving Ground Real World

slide-19
SLIDE 19

19 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

Toolchain Validation Example Vehicle Dynamics

“The validity of the applied modelling and simulation tool shall be verified by means of comparisons with practical vehicle tests. The tests utilized for the validation shall be the dynamic maneuvers (…)“

Source: Uniform provisions concerning the approval of passenger cars with regard to Electronic Stability Control (ESC) Systems, ECE R140

Correlation Results and Approval of Simulation Usage of Simulation for ADAS/AD Validation Determination and Identification of Parameters Parameterization of Simulation Models

ADAS/AD Validation Mandatory Preparation (Validation and Approval of Models and Toolchain)

slide-20
SLIDE 20

20 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

Toolchain Validation Example Vehicle Dynamics

Vehicle prototypes Virtual prototype

slide-21
SLIDE 21

21 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

Challenge Co-Simulation (1)

Boundary Conditions: In the area of simulation there is not only ONE tool

  • r ONE model. Different tools have different

advantages, different models are for different use cases (e.g. dynamic scenarios require complex vehicle dynamics) Example bandwidth based on customer use cases … and the landscape is broad!

# Powertrain Vehicle Dynamics Environment Function 1 CUS Simulink MSC ADAMS VIRES VTD CUS FMU 2 AVL VSM AVL VSM CUS Tool ROS1 Node 3 Real on Test Bed Real on Test Bed VIRES VTD On control unit 4 CUS Simulink CUS Simulink IPG CarMaker ROS2 Node

Requirements and Challenges (Overview): An open integration platform is crucial! The platform need to combine simulation and real components up to the real vehicle Integration on simulation level is more the exchanging signals!

slide-22
SLIDE 22

22 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

Challenge Co-Simulation (2)

Requirements and Challenges (Details):

Technical View: ▪ Multi-domain development ▪ Multi-tool approach ▪ Multi-vendor ▪ Dynamic coupling ➢ Virtual prototype representation

Example:

AEB Scenario: “Full braking after acceleration to 100 km/h” k, m, n >>

solver solver solver solver solver solver solver

Mathematical View: ▪ Multi-method ▪ Multi-solver ▪ Multi-rate ▪ Dynamic coupling ➢ Coupling error Significantly longer braking distance (~1.9 m) due to coupling error! Correct (co-)simulation result with NEPCE*) in Model.CONNECT™

*) NEPCE … Nearly Energy Preserving Coupling Element

slide-23
SLIDE 23

23 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

Implementation with NVIDIA Tools Reference: Proving Ground

NVIDIA DRIVE AD Function Vehicle Interface Real Vehicle on Real Proving Ground Real Vehicle on Real Proving Ground Methodology, Test Catalogue and KPIs for AD Function AVL DRIVE for ADAS KPI SC Vehicle Under Test

Proving Ground

very close to the real

  • peration

expensive, high effort, less repeatability

slide-24
SLIDE 24

24 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

Implementation with NVIDIA Tools Virtual Testing @ the AVL DRIVINGCUBE™

NVIDIA DRIVE AD Function Vehicle Interface CAN Interface TB Interface AVL VECON Real Vehicle on AVL DRIVINGCUBE™ Virtual Vehicle on Virtual Proving Ground Testbed.CONNECT™ with Model.CONNECT™ NVIDIA DRIVE SIM DRIVE SIM Wrapper Methodology, Test Catalogue and KPIs for AD Function AVL DRIVE for ADAS DRIVE Interface KPI SC Vehicle Under Test

Virtual Testing

close to real operation, chassis dynos are already established for homologation (emissions) limited in terms of lateral dynamics

slide-25
SLIDE 25

25 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

slide-26
SLIDE 26

26 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

…as vehicle integration lab

Different functions and perceptions must be evaluated at a certain time in the vehicle. The DRIVINGCUBE is the only test environment for reproducible and repeatable test on vehicle level!

… for security testing

Hacking attacks to evaluate the security of the vehicle must also be performed during operation

  • f the vehicle.

Do you want to try an attack on a highway at high speed?

… to reproduce critical scenarios and tests

Critical Scenarios (in general or determined

  • ut of simulation) must

be analyzed on vehicle level in a reproducible way.

… as most efficient test instance for a lot of use cases

Different uses cases in validation and also homologation can not be executed efficiently in other test instances (e.g. ECE 79: LKA above 130km/h)

The AVL DRIVINGCUBE™…

slide-27
SLIDE 27

27 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

…as vehicle integration lab: Details

Steering Braking Vehicle Perception Autopilot long HMI Powertrain Autopilot lateral Thermal

DRIVINGCUBE as vehicle integration lab

slide-28
SLIDE 28

28 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

…as vehicle integration lab: Details

Steering Braking Vehicle Perception Autopilot long HMI Powertrain Autopilot lateral Thermal

DRIVINGCUBE as vehicle integration lab

The AV functions must always be validation in combination with the vehicle (dynamics, performance, behavior) Not each detail can be modelled and provided for simulation

slide-29
SLIDE 29

29 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

…as vehicle integration lab: Use Case / Example

Vehicle: T Vehicle: D

In a study with one of our research partners we figured out that even in one model series (so nearly the same vehicle – Vehicle T and D) a LKA function behaves completely different and the calibration of this function had to be adapted.

KPI: Curve-Cutting-Gradient (CCG)

▪ Calculate Distance to centerline (D2L) by using Ground-Truth-Maps ▪ Evaluate (de)position of VUT during driving with LKAS

Mean Distance to centerline [m] Amount [-]

Evaluation of a Lane Keep Assist:

slide-30
SLIDE 30

30 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

Implementation with NVIDIA Tools Architecture Simulation

NVIDIA DRIVE AD Function CAN Interface VSM Wrapper AVL VSM Virtual Vehicle in Vehicle Simulation Virtual Vehicle on Virtual Proving Ground Model.CONNECT™

  • NVIDIA

DRIVE SIM DRIVE SIM Wrapper Methodology, Test Catalogue and KPIs for AD Function AVL DRIVE for ADAS DRIVE Interface KPI SC Vehicle Under Test

Simulation

fast, flexible and cheap in

  • peration, there are already

ESC homologation processes in simulation is

  • nly as good as the

model(s)

slide-31
SLIDE 31

31 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

First Results / First Comparison

5 10 15 20 25 30 35 40 45 50 55 60

Time / s Velocity TSV / (m/s)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Proving Ground VHiL Repetition 1 MiL 5 10 15 20 25 30 35 40 45 50 55 60

Time / s Desired Acceleration / (m/s^2)

  • 3,0
  • 2,5
  • 2,0
  • 1,5
  • 1,0
  • 0,5
0,0 0,5 1,0 1,5 2,0 2,5 3,0 Proving Ground VHiL Repetition 1 MiL 5 10 15 20 25 30 35 40 45 50 55 60

Time / s Delta Distance / m

5 10 15 20 25 30 35 40 45 50 Proving Ground VHiL Repetition 1 MiL 5 10 15 20 25 30 35 40 45 50 55 60

Time / s Velocity / (m/s)

  • 2
2 4 6 8 10 12 14 16 Proving Ground VHiL Repetition 1 MiL upper & lower limit

Riedmaier, S.; Nesensohn, J.; Gutenkunst, C.; Düser, T.; Schick, B.; Abdellatif, H.: Validation of X-in-the-Loop Approaches for Virtual Homologation of Automated Driving Functions, GSVF Symposium 2018, Graz

slide-32
SLIDE 32

32 Public Tobias Dueser | AVL List GmbH | 2019-03-20 |

Validation and Homologation of Autonomous Driving will be a challenge Partnering is extremely important to join the forces The most efficient validation and homologation will be done by those who will use the best combination of different test environments NVIDIA, TUEV SUED AND AVL will push this topic together