Automated Driving & Development Processes – A ...

169 downloads 0 Views 3MB Size Report
Nov 9, 2017 - Dr. Olaf Schädler, Sven Hallerbach (Opel Automobile). Matthias Büker , Sebastian Gerwinn (offis),. Tino Brade (Robert Bosch), and Christian ...
Automated Driving & Development Processes – A brief look beyond Dr. Ulrich Eberle (Opel Automobile) Dr. Olaf Schädler, Sven Hallerbach (Opel Automobile) Matthias Büker , Sebastian Gerwinn (offis), Tino Brade (Robert Bosch), and Christian Amersbach (Darmstadt University – FZD)

Aachen, 9th November 2017

The Traditional Starting Point Sensor Configuration

Different Sensor Views

Sensor Fusion Planning and Decision Making in an Interpretation of the Physical World

A Look Beyond… New challenges emerging due to high situational complexity

Event-triggered TL Path B

Path Y1

Path V1

Path Y2

Path B Vehicle 1

Optimized Approach

Unknown Intentions

Communication Infrastructure

Traffic Environment Missing information on Driver Intentions Complex Road Structure Signal Phases & Queues

Not even considering pedestrians, cyclists, kids etc.

Difficult Structures

Vehicle 2

Exemplary Test Approach for Automated Driving – Coupling of VEHICLE DYNAMICS and TRAFFIC FLOW simulations using DIGITAL TWIN

Hallerbach S., Eberle U., Köster F. : Absicherungs- und Bewertungsmethoden für kooperative und hochautomatisierte Fahrzeuge. AAET - Automatisiertes und vernetztes Fahren. Braunschweig, (Research Paper). 8.-9.02.2017

But there are simpler tasks than „just“ creating such a Digital Twin …

V2X/ITS G5

4G 5G

And Finally Facing Again the Old Enemy: Complexity ACC: Long-Range-Radar and ECU:

O. Schaedler (Opel)

 36 Possible Combinations

ACC and LCA: Long-Range-Radar, Camera and ECU:

O. Schaedler (Opel)

 288 Possible Combinations

An Exemplary Development Process Vehicle w/o Automation (early 2010s) and w/o considering Powertrain Ca. 400 Manufacturing Options, 30 Electronic Control Units, Several Hundred Cal-Files, Thousands of Parameters

Duration: Several years, depending on specific OEM

Start of Regular Production

Level 1

Level 2

Level 3 T. Tiecke et al. (Opel)

V Model

9

V Model and ISO 26262

10

ISO and SOTIF

T. Weispfenning (Opel)

ISO26262 only as basis … „Safety of the Intended Functionality“ needs to be covered as well. NEW CHALLENGES: Mix Traffic and Issues at Other Cars 11

Process Extension

plus contingently - „Silent Testing“ Approaches - Staggered Deployment via Restricted Fleets as well as the corresponding feedback cycles

12

Exemplary Impact on OEM Development

A. Weitzel et al. (Opel) 13

Exemplary Impact on OEM Development

14

What is the impact on PEGASUS methodology?

 PEGASUS method needs to be coordinated with established OEM and industry processes  Do we miss something when applying a scenario-based approach  How may the testing effort be structured?

15

Approach to Assess and Structure Challenges

16

Specific Example : Information Basis

17

Approach to Structure Test Effort Functional Decomposition  A scenario-based approach is able to reduce the test effort 

BUT the various scenarios still check all functional levels simultanously.



High number of combinations of initial and boundary conditions lead a large number of test cases

 The functional decomposition and the structuring of relevant scenarios into „test case quants“, which test in structured way just one (or a few) functional levels: 

Reduces the overall test effort



Enables the application of tools (XiL, proving ground etc.) according to their validity.



A reduction of the test risk.

18

Functional Decomposition Relevant Scenarios (functional) Matrix Definition and Assignment of Failure Criteria via FTA

Elimination and Bundling of Failure Criteria Decision on Appropriate Test Cases and Test Tools for Remaining Failure Criteria

Scenarios Functional Levels

Which requirements result from certain scenarios for specific levels

19

Outlook  New development approaches should be as „downward-compatible“ to the existing automotive process world as possible (and as making sense)  AD vehicles initially could have a lower diversity than current cars to cope with complexity issues  Structured approaches needed to make clear what is known, what is not known and about what not even is known to be unknown.  Structured approaches like functional decomposition needed to reduce test efforts  To introduce AD cars, concepts like „silent testing“ in restricted fleets and a staggered deployment to real-world application including learning cycles are worth to be examined

 PEGASUS is working on these challenges

20

Thank you … Merci beaucoup

21

What will change for Automated Driving?  Robust Sensor Technology over Vehicle Lifetime  Fail-operational Systems  Integration of Connectivity Concepts (V2X) Source Wikipedia

How to cope with these challenges? •

Role of Simulation Tools and Test Automation



Documentation Technological State-of-the-Art



Definition of a Safe State



Role of Technology Redundancy



Financial Viability / Business Model