ADAS/AD Challenge. Copyright 2017, dspace GmbH

Transcription

1

2 ADAS/AD Challenge 2

3 dspace Automotive Simulation Models (ASM) for ADAS and AD Michael Peperhowe, Group Manager ASM VD & Traffic dspace GmbH Rathenaustr Paderborn Germany 3

4 ASM Overview 4

5 ASM Automotive Simulation Models The brain of the simulation ASM - Automotive Simulation Models 5

6 ASM Multi Domain ASM - Automotive Simulation Models Multi Domain Tool Suite Drivetrain Electrification Vehicle Dynamics ADAS / AD Open Simulink Real-Time Network 6

7 ASM Tool Suite Simulation Parameterization Animation 7

8 ASM Multi Platform Unit test (MIL) Unit test (SIL) System test (HIL) System test (XIL) Vehicle test (real world) ML/SL VEOS RCP HIL Test Bench VIL 8

9 ASM for ADAS & AD 10

10 Use-Case Complex City Scenario Application ASM Traffic Road Traffic Sensors Scenario >120 detectable objects Complex scenery (buildings, busstops, roadworks, ) Sensors Multiple 2D & 3D-object sensors with realistic sensor geometry Soft ECU ACC + AEB MotionDesk 12

11 ASM Road 13

12 ASM Road Key Features Road networks consisting of roads and junctions Independent height and surface definitions Lanes with smooth transitions and specific line definitions New 18-A: Unlimited number of lanes Easy-to-use interactive road editor in ModelDesk Data conversion from real world road data 14

13 Freely Definable Shapes and Trajectories for Realistic Sceneries Continuous objects e.g. concrete barriers Free trajectory e.g. drive in road works Repeating objects e.g. road barriers, lining poles NEW 2016/2017 Free lines (customizable color and style) 15

14 ASM Road Conversion from Real Data Road conversion Real intersection BLOM OpenDRIVE import and export, OpenCRG import GPS or X,Y,Z measurement data OpenStreetMaps, ADAS RP, Here, SHAPE, Google Earth, HAD maps (*) (*) in the context of projects 16 - Confidential

15 ASM Road HAD Map Data Import HAD map data High accuracy road description (absolute, relative) Lane dividers, road borders, guard rails Lane dependent attributes, e.g. speed limits Currently available: German Autobahn, French Highways, US Interstates 17 - Confidential

16 ASM Road HAD Example Data German A33/A2 (close to Paderborn) Conversion of Horizontal profile Vertical profile (incl. bridges) Lanes (incl. accelerating and decelerating) Borderlines / Emergency Lanes TomTom data (shape file) JOSM import 18 - Confidential

17 German Autobahn A33/A2 Bing 19 - Confidential

18 ASM Traffic 20

19 ASM Traffic Scenario Definition Easy and intuitive scenario definition (UML-like) Improved flexibility and usability Reuse of configured scenarios in different projects by copy/paste Independent definition of fellow scenarios Start Activity Segment Transition Flexible start point Individual activity length Individual activity transition Unlimited number of fellows Pre-defined scenarios for Euro NCAP OpenScenario (OSC) core team 21

20 Ego vehicle Fellow ASM Traffic Import of GIDAS Data GIDAS: German In-Depth Accident Study PCM: Pre-Crash Matrix Contains reconstruction data from the accident ( digital accident sketch ) Contains data from the collision of the first two accident parties Coordinate system s origin is the first collision point VUFO GmbH 22 - Confidential

21 ASM Traffic GIDAS Example Scenario Original data Scenario with AEB 23 - Confidential

22 Traffic Flow Simulation Features Simulation of realistic traffic flow (microscopic simulation i.e. movement of single vehicles) Evaluation of infrastructure changes Remote control via API possible SUMO - Simulation of Urban Mobility Free of charge and Open-Source Developed by DLR (German Aerospace Center) Support of V2X Communication (e.g. Veins) OpenStreetMaps as import format PTV Vissim Commercial tool Developed by German PTV Group Source: PTV 24 dspace-confidential

23 ASM Traffic Co-Simulation with SUMO & Vissim Under Development ASM ASM Ego Position Ego Traffic flow simulation Fellows Ego Veh 1 Veh 2 ASM Sensors Communication Interface Veh m Veh 3 Veh m+1 Position Fellows ASM Fellows Extrapolation Veh n Veh n 1 Δt = 0.001s Δt = 0.1s PTV n Vissim n ASM 25 - Confidential

24 ASM Traffic Co-Simulation with Vissim

25 ASM Smart Traffic Under Development ASM ego vehicle and fellows Recognition of Priority signs Traffic lights Speed limits Stop signs Collision avoidance with other fellows Use-cases Random traffic for robustness tests of ADAS/AD RDE with realistic traffic simulation 27

26 ASM Smart Traffic Example Scenario 28

27 ASM Smart Traffic Example Scenario 29

28 ASM Multi-Agent Simulation Multiple vehicles with complex vehicle dynamics (MBS, tire, suspension, steering, ) Multiple vehicles with sensors One vehicle per VPU (VEOS) or core (SCALEXIO) Use-Cases: AD Platooning V2V communication Interactive driving simulators 30

29 ASM for NCAP Scenarios Euro NCAP LSS: Lane Support Systems Specification of free trajectories possible as of release 2017-B Exact definition of lane markers (type, color, width, etc.) Euro NCAP SAS: Speed Assist Systems New traffic sign sensor as of release 2017-B Traffic lights, variable message signs, lane dependency, OpenDRIVE coding, etc. Euro NCAP AEB and AEB VRU: Automatic Emergency Braking Systems AEB test scenarios readily available in provided demos Project solution for AEB VRU View-Obstruction in 3D sensor Triggered pedestrian/bicyclist movement for exact collision position and timing Vehicle front impact profile definition for exact collision detection and test evaluation 31

30 Sensor Simulation 32

31 Sensor Models dspace Solutions Reality Physical Real Lens Test Bench Over the Air (OTA) Imager Image processing Phenomenological MotionDesk GPU-based Lens Imager Image processing Probabilistic Ideal Ground Truth ASM Complexity 33

32 ASM Ground Truth Sensors 34

33 ASM Sensor Updates Updated 2016/2017 Unlimited number of sensors 3D sensor with realistic sensor scope Range, cone and death zone length Vertical, horizontal radius Sensor scheduling 2D sensor with free space detection Array of sensor rays in the x-y plane Detection of guard rails Traffic sign sensor for traffic lights and gantries Detection of gantries incl. related road lane Advanced OpenDrive interface for import/export of traffic signs 35

34 Use Case Camera Sensor Emulation AEB (Automatic Emergency Braking) ASM 3D Sensor LKA (Lane Keep Assist) ASM Line Sensor ISA (Intelligent Speed Assist) ASM Traffic Sign Sensor 36

35 ASM for Open Simulation Interface (OSI) Under Development Research project PEGASUS Effective validation of ADAS/AD systems Audi, Bosch, BMW, Conti, Daimler, VW, E.g. standardization of sensor interface Decoupling of environment and sensor models Unified description of environment data via Google Protocol Buffer (GPB) ASM global data for objects and road information, like lanes Environment\ Traffic OSI Data Sensor1 OSI Sensor2 OSI 37

36 MotionDesk Sensors GPU Based 38

37 MotionDesk Camera Sensor (RLS 2016-B) NEW Simulation of lens distortion effects: Radial distortion Chromatic aberration Vignetting Calibration from real lens-camera combinations Adobe camera lens profile Adobe lens distortion model Configuration at MotionDesk 39

38 MotionDesk 3D Point Cloud (RLS 2017-A) NEW Ideal simulation of laser sensors or ray-based systems Detection of all elements within the virtual sensor at MotionDesk Feedback of the information of all detected points into ESI, RTMaps or VEOS VPU, e.g. the distance to the objects and relative velocity (*) Freely configurable field of view up to 180 and resolution, e.g points with 60 fps Closed-loop tests of ADAS applications incl. the image processing unit on a standard PC MotionDesk (*) in context of engineering projects

39 MotionDesk - Advanced Lighting Technology OUTLOOK 2017-B ff. RLS 2017-B: Realistic lighting behavior Shadow casting Unlimited number of light sources e.g. street lamps and headlights RLS 2018-A/B: High-dynamic range (HDR) support Dazzling effects, glare, lens flares, reflection on roads will be gradually implemented (starting from 18-A) 41

40 Conclusion 42

41 Conclusion An open and continuous simulation environment speeds-up your ADAS/AD development. Integration of real world data is necessary and possible. Sensor simulation on all stages gives new possibilities. ASM brings your real world to the virtual environment. ADAS/AD is a big challenge. But you can beat it with the right tooling! MPeperhowe@dspace.de 43

42 ADAS/AD Beat the Challenge 44

43 Thank you for listening! 45

44 Important Information! 2017, dspace GmbH All rights reserved. Written permission is required for reproduction of all or parts of this publication. The source must be stated in any such reproduction. This publication and the contents hereof are subject to change without notice. Benchmark results are based on a specific application. Results are generally not transferable to other applications. Brand names or product names are trademarks or registered trademarks of their respective companies or organizations. 46

45