Using Virtualization to Accelerate the Development of ADAS & Automated Driving Functions GTC Europe 2017 Dominik Dörr 2
Motivation Virtual Prototypes Virtual Sensor Models CarMaker and NVIDIA DRIVE PX 2 Conclusion 3
USING VIRTUALIZATION TO ACCELERATE THE DEVELOPMENT OF ADAS & AUTOMATED DRIVING FUNCTIONS Motivation Virtual Prototypes Virtual Sensor Models CarMaker and DRIVE PX 2 Conclusion 4
Safety-Critical Functions Must Not Fail! Testing in countless everyday situations Millions / billions of km necessary to release functions in real-world tests Changes to the software (code or parameters) Start testing again! 5
USING VIRTUALIZATION TO ACCELERATE THE DEVELOPMENT OF ADAS & AUTOMATED DRIVING FUNCTIONS Motivation Virtual Prototypes Virtual Sensor Models CarMaker and DRIVE PX 2 Conclusion 6
Automotive Systems Engineering in the Development Process Providing internal departments + suppliers with virtual prototypes Internal dept. 1 or supplier 1 Component A Internal dept. 2 or supplier 2 Component B Internal dept. 3 or supplier 3 Component C Internal dept. N or supplier N Component Z Virtual Prototypes Integration workshop Internal development depts. or suppliers Development + test of full system in full vehicles in scenarios Pre-calibration of the system Test driving department Calibration (virtual + real) Release (also virtual) Engineers can test and release their systems in a full vehicle: automotive systems engineering! 7
Automotive Systems Engineering in the Development Process Using virtual prototypes throughout the entire development process Same maneuvers + same evaluation criteria Reuse of virtual prototypes and maneuvers saves time and costs 8
Test Scenario Definition Test scenario Static objects Traffic signs, buildings, road incl. markings,... Virtual prototype AD functions, (sub-) system models, driver,... Dynamic objects Cars, trucks, pedestrians,... Surrounding of virtual prototype Environmental conditions Lighting, temperature, precipitation,... 9
USING VIRTUALIZATION TO ACCELERATE THE DEVELOPMENT OF ADAS & AUTOMATED DRIVING FUNCTIONS Motivation Virtual Prototpyes Virtual Sensor Models Radar as Example CarMaker and DRIVE PX 2 Conclusion 10
Environment Function / fusion Use Case-Specific Sensor Models Overview The use case drives the level of detail: Ideal Sensors for rapid prototyping / proof of concept HiFi Sensors for function development & testing Raw Signal Interface for component / signal processing development & testing Virtual scenario Ideal Sensor Virtual prototype with system under test HiFi Sensor Raw Signal Interface Processing / tracking Use case-specific sensor models for real-time simulation! 11
Do the algorithms work? Requirements: Function interface (e.g. object list) Technology-independent Easy to parameterize Does the function generally work? Requirements: Function interface (e.g. object list) Technology-specific Physical phenomena Processing/Target selection included Does the sensor component work? Requirements: Raw signal interface Technology-specific Detailed physical effects 12
Suitable Sensor Models for Every Use Case Ideal Sensors for verification and early development phases Scenario Function / Fusion Ideal Sensor Information extraction Object selection HiFi Sensor Propagation reduced to essentials Target selection RAW Signal Interface Signal propagation Processing / tracking 13
Object and Line Sensor Highlights Object sensor detects and tracks surrounding traffic Object list Relevant target Line sensor generates object list of lane boundaries Road markings Traffic barriers Application areas ACC, LDW, LKA, 14
Radar Sensor Outputs: Object lists with Relative position / velocity / acc Classification of object / dynamics / confidence Probability of existence / obstacle Received power / RCS / SNR Features: Detection based on SNR including occlusion Antenna gain Object RCS Propagation loss Object fusion based on resolution cell Latency, noise False negatives (False positives) 15
16
Radar RSI Preview Effects: Multipath propagation Repeated path echo Coherent addition Doppler shift False positives/negatives 17
18
USING VIRTUALIZATION TO ACCELERATE THE DEVELOPMENT OF ADAS & AUTOMATED DRIVING FUNCTIONS Motivation Virtual Prototpyes Virtual Sensor Models CarMaker and DRIVE PX 2 Conclusion 19
CAN Data Coupling of CarMaker and DRIVE PX 2 Platform OR Visualization Using IPGMovie as visualization CarMaker Visualization IPGMovie Automotive Camera films video output from Video Interface Box monitor Sample algorithms from Driveworks on DRIVE PX 2 platform do lane recognition or Control Signals DRIVE PX 2 object detection Closing the loop is possible DRIVE PX 2 platform as prototyping ECU solution Closing the loop 20
USING VIRTUALIZATION TO ACCELERATE THE DEVELOPMENT OF ADAS & AUTOMATED DRIVING FUNCTIONS Conclusion Virtual prototypes can be used throughout the whole development process Virtual sensor models provide the virtual prototype with information about the scenario In different levels of detail for different use cases DRIVE PX 2 can be included in the closed-loop simulation process Virtual testing is crucial for ADAS & Automated Driving functions IPG Automotive offers the platform 21
China France India Italy Japan Korea Malaysia Sweden Turkey UK USA Headquarters Germany: Bannwaldallee 60 76185 Karlsruhe Tel.: +49 721 98520-0 ipg-automotive.com