David Howarth Business Development Manager Americas
David Howarth IPG Automotive USA, Inc. Business Development Manager Americas david.howarth@ipg-automotive.com ni.com
Testing Automated Driving Functions using a flexible XiL Approach IPG background: Founded in 1984; pioneer in vehicle dynamics simulation Global expertise in Virtual Vehicle development & simulation Industry leading CarMaker Software for ADAS & AV
Safety-Critical Functions Must Not Fail! Testing in countless everyday situations Changes to the Software Start testing again! Millions+ of miles necessary to release functions in real-world tests 23-May-18 6
Virtual Vehicle Development & Simulation is the only Option
Virtual Vehicle Development & Simulation is the only Option
Using XiL to test Autonomous Vehicle Functions Combine XiL with Virtual Vehicles and Simulation Use Virtual Vehicles & Simulation from the beginning of development Virtual Vehicle Prototype Test AV functions in a representative functional vehicle model, not in isolation Higher quality results from the beginning
Using XiL to test Autonomous Vehicle Functions Enable Reuse across the entire Development Process Virtual Vehicle Prototype Hardwarein-the-Loop Vehiclein-the-Loop Model-inthe-Loop Softwarein-the-Loop Processorin-the-Loop Same simulation scenarios + evaluation criteria
Using XiL to test Autonomous Vehicle Functions What is needed to add Virtual Vehicles and Simulation? 1. Utilize a platform to create high-fidelity vehicle models with sensors 2. Build simulation scenarios with: roads, traffic, and driver 3. Create or modify simulation scenarios quickly GUI based 4. Control traffic vehicles and pedestrians 5. Work with other tools (e.g. NI) 6. Run simulations in faster than real-time 7. Utilize test automation, scripting, and DOE (Design of Experiments)
Virtual Vehicle Development A complete vehicle testing environment for real-world driving simulation Comprehensive vehicle model with integrated powertrain models Easy exchange of 3 rd party models for vehicle sub-systems (e.g., Simulink, FMI, C-code, etc.)
Virtual Vehicle Development Complete Realtime Vehicle Model Aerodynamics 3D aerodynamics Crosswind build-up Vehicle Body Flexible or rigid Sprung mass, loads, engine Position, masses, inertia Steering Steer by torque / angle Pfeffer steering model Sensors Ideal Sensors High-Fidelity Sensors Raw Signal Interface Sensors Powertrain Engine (Layouts, Mappings) Elastic driveline shafts Clutch (Friction, Converter) Gearbox (automatic / manual) Driveline Suspension Spring (Linear, Non-linear) Damper (Linear, Non-linear) Buffer (Linear, Non-linear) Stabilizer Kinematics & Compliance Wheel bearing with friction Tires TameTire and Pacejka 3D Tire Interface, MF-5.2, MF-Swift etc. Brakes Hydraulic model valid for ABS/ESP Pressure distribution via pedal actuation or via pedal force
Where do You get a Virtual Vehicle? Practical vehicle parameterization based on publicly available information
Virtual Vehicle Development Sensors: Specific models for a wide range of use cases Sensor Model Type Typical Use Case Real-Time Functionality Sensor Model Output Physical Sensor Effects IDEAL Function Proof-of-Concept Yes Object List (FS+ Sensor provides Point Cloud) None Need a wide range of sensor models to support specific testing needs 23-May-18 15
Virtual Vehicle Development Sensors: Ideal Sensor Need a wide range of sensor models to support specific testing needs 23-May-18 16
Virtual Vehicle Development Sensors: Specific models for a wide range of use cases Sensor Model Type Typical Use Case Real-Time Functionality Sensor Model Output Physical Sensor Effects IDEAL Function Proof-of-Concept Yes Object List (FS+ Sensor provides Point Cloud) None HIGH FIDELITY Function Development Yes Object List Extensive Need a wide range of sensor models to support specific testing needs 23-May-18 17
Virtual Vehicle Development Sensors: Specific models for a wide range of use cases Sensor Model Type Typical Use Case Real-Time Functionality Sensor Model Output Physical Sensor Effects IDEAL Function Proof-of-Concept Yes Object List (FS+ Sensor provides Point Cloud) None HIGH FIDELITY Function Development Yes Object List Extensive PHYSICAL Sensor Development Yes (depends on complexity & computing capacity) Raw Signal Full Need a wide range of sensor models to support specific testing needs 23-May-18 18
Virtual Vehicle Development Example: Radar Raw Signal Interface
Virtual Vehicle Development A complete vehicle testing environment for real-world driving simulation Comprehensive vehicle model with integrated powertrain models Easy exchange of 3 rd party models for vehicle sub-systems (e.g., Simulink, FMI, C-code, etc.) Fast and easy import of detailed real-world road networks GUI construction of roads
Virtual Vehicle Development Roads: Import, or Measure Realistic Road Networks Here, Google Maps ADASRP, KML Measurement-based High-fidelity mapping services Source: 3D Mapping 23-May-18 21
Virtual Vehicle Development GUI Drag & Drop Tool to create Roads+ Tool-bar on the left Roads, Intersections Streetsigns Houses, Trees, Bridges
Virtual Vehicle Development GUI Drag & Drop Tool to create Roads+
Scenario Generation
Virtual Vehicle Development A complete vehicle testing environment for real-world driving simulation Comprehensive vehicle model with integrated powertrain models Easy exchange of 3 rd party models for vehicle sub-systems (e.g., Simulink, FMI, C-code, etc.) Fast and easy import of detailed real-world road networks GUI construction of roads Intelligent driver model with definable driver attributes Replaceable with AV control models Directed and autonomous traffic objects Large variety of traffic objects to create high complexity scenarios
Virtual Vehicle Development Traffic: Create, Generate, Co-Simulate, or Record / Import Create Manually Generate Randomly Import Recorded Traffic Front view Rear view Traffic Editor Fully parameterizable, including radar cross sections Defined, directed, & autonomous driving modes with lateral motion Traffic Generator Part of Scenario Generator Definable density and vehicle types Place hundreds of traffic items in seconds Record, Replay, Rearrange Record & annotate video Reconfigure traffic & driver behavior to set up comprehensive test catalog 23-May-18 26
Virtual Vehicle Development Traffic Capabilities Drive arbitrary paths Speed profiles Autonomous driving Physics-based with roll and pitch Pedestrians with natural motion Animals with natural motion
Virtual Vehicle Development Dangerous Maneuvers Safely generate test-cases that are very difficult or dangerous Important to test dangerous maneuvers early in the development process
Using XiL to test Autonomous Vehicle Functions Enable Reuse across the entire Development Process Virtual Vehicle Prototype Hardwarein-the-Loop Vehiclein-the-Loop Model-inthe-Loop Software-inthe-Loop Processorin-the-Loop Same simulation scenarios + evaluation criteria
Example Use Case: MiL Co-Simulation with Simulink PC Direct Interface for co-simulation CarMaker Simulink Function under test HDMI Focus on the model behavior while working in Simulink Easily test the model as part of a Virtual Vehicle running in CarMaker Interactively or automated 30
Example Use Case: MiL C/C++ Based Models For organizations who build function models in C/C++ Integrate C/C++ models into CM POC, function prototypes PC CarMaker HDMI
Example Use Case: MiL ROS: Robot Operating System ROS is widely used for AV function development and POC Node 1 Node 2 ROS + CarMaker for virtual vehicle development Simulate the missing Vehicle environment Co-simulation: timing synchronization, closed-loop Straightforward data exchange between CarMaker and ROS
Using XiL to test Autonomous Vehicle Functions Enable Reuse across the entire Development Process Virtual Vehicle Prototype Hardwarein-the-Loop Vehiclein-the-Loop Model-inthe-Loop Softwarein-the-Loop Processorin-the-Loop Same simulation scenarios + evaluation criteria
Example Use Case: SiL Auto-code w/ Simulink to test C functions Simulink Function PC CarMaker C-code Validate behavior of auto-generated code The focus is on testing of the SW Code-coverage, types, etc. Enables back-to-back tests C-code HDMI
Example Use Case: SIL Using Functional Mock-up Interface 2.0 Share SW functions to/from suppliers without sharing the IP Seamless and easy integration of own or third-party models into CarMaker
Example Use Case: SiL HPC: High Performance Computing High Performance Computing (HPC) Cluster testing on remote servers Run thousands to millions of simulations in parallel GUI IPGMovie Custom APO Client Manage the test execution and results Host PC This is the mission-critical capability needed for AV SW validation SimCore 1 SimCore 2 SimCore 3 Supports 1 M+ mile overnight testing, to validate AV SW changes Run the same Virtual Vehicle and simulations already created Computer A SimCore 65 SimCore 66 SimCore 67 Use DOE approach to quickly create more tests Computer B Communication via network
Using XiL to test Autonomous Vehicle Functions Enable Reuse across the entire Development Process Virtual Vehicle Prototype Hardwarein-the-Loop Vehiclein-the-Loop Processorin-the-Loop Model-inthe-Loop Software-inthe-Loop Same simulation scenarios + evaluation criteria
Example Use Case: PiL Test Video Perception + Path Planning + Vehicle Control on Embedded HW PC USB USB to CAN Adaptor CAN ECU e.g. Nvidia Drive PX2 CarMaker VIB DVI GMSL Function under test HDMI AV performance testing on representative HW E.g. testing Neural-Net object detection algorithm on automotive embedded HW Test for faults generated by target compiler
Using XiL to test Autonomous Vehicle Functions Enable Reuse across the entire Development Process Virtual Vehicle Prototype Hardwarein-the-Loop Vehiclein-the-Loop Model-inthe-Loop Software-inthe-Loop Processorin-the-Loop Same simulation scenarios + evaluation criteria
Example Use Case: HiL PC TCP/UDP NI PXI RTPC CarMaker CAN, FlexRay, LIN, TCP/UDP ECU e.g. Nvidia Drive PX2 IPG Movie DVI VIB GMSL Function under test HDMI
CarMaker Runs on Different Hardware Platforms Function under test integrated in CarMaker HIL software as middleware running on Hardware platform (National Instruments PXI is fully supported)
CarMaker/HIL on National Instruments Hardware Fusion of leading technologies CarMaker Platform PXI Platform Supported NI SW IPG HW for NI FPGA-based IPG M- Module Fully PXI compatible Customized FPGA Sensor Simulation Camera Simulation NI PXI or NI PXIe realtime system LabVIEW Real-Time Module NI-CAN for direct CAN Access NI-XNET NI-VISA for IPG HW and USB access LabVIEW NI VeriStand VeriStand NI VeriStand Engine
Using XiL to test Autonomous Vehicle Functions Enable Reuse across the entire Development Process Virtual Vehicle Prototype Hardwarein-the-Loop Vehiclein-the-Loop Model-inthe-Loop Software-inthe-Loop Processorin-the-Loop Same simulation scenarios + evaluation criteria
Example Use Case: ViL 1. All embedded SW is running on a physical vehicle ECU(s) 1. PC is running Simulation scenarios to bypass physical vehicle sensor(s) 2. Test a physical vehicle on a real road, but with sensor input from simulation 1. Emergency braking today companies push test-dummies in front of moving cars 3. Enables final physical validation with minimal physical prototypes, leveraging the Simulation Scenarios
Example Use Case: ViL with Video Injection
Using XiL to test Autonomous Vehicle Functions Enable Reuse across the entire Development Process Virtual Vehicle Prototype Hardwarein-the-Loop Vehiclein-the-Loop Model-inthe-Loop Softwarein-the-Loop Processorin-the-Loop Same simulation scenarios + evaluation criteria
Using XiL to test Autonomous Vehicle Functions Enable Reuse across the entire Development Process Virtual Vehicle Prototype Same simulation scenarios + evaluation criteria
Use Virtual Vehicle Development across Domains What is needed to implement AV Virtual Vehicle Development for XiL? 1. Develop with a representative vehicle (virtual) from Day #1 1. Front-load development: reduce bugs and find them earlier 2. Share Virtual Vehicles and Simulation Scenarios across the V-cycle and different development domains 1. Requires a Vehicle Model database and Test Catalog 3. HPC is required to run the 1 M+ tests for releasing AV functions 1. Reuse the existing Virtual Vehicle and simulations 4. Higher quality, more mature SW earlier!
Testing automated driving functions using a flexible XIL approach Thank You!