Auto electronics
Informationalized Solution
R&D tools
Marketing
From Functionality to Safety: Building a Multi-dimensional Quality Assurance System
The autonomous driving controller serves as the "brain" of the intelligent driving system. It fuses data from various sensors—including cameras, LiDAR, millimeter-wave radar, and ultrasonic sensors—to perform environmental perception, decision-making, planning, and vehicle control. This controller directly determines the safety and reliability of autonomous driving functions.
Perception Fusion: Processes multi-sensor data from cameras, LiDAR, millimeter-wave radar, and other sensors.
Decision Planning: Conducts path planning and behavioral decision-making based on environmental perception results.
Vehicle Control: Outputs control commands such as throttle, braking, and steering.
Safety Monitoring: Monitors system status in real-time and executes functional safety strategies.
ADAS/AD Controller Sensor Layout Diagram
GB 39901-2025: Technical requirements and testing methods for Advanced Emergency Braking Systems (AEB) of light vehicles.
Intelligent and Connected Vehicles — Combined Driver Assistance System Safety Requirements (Draft for Comment).
Intelligent and Connected Vehicles — Automated Driving System Safety Requirements (Draft for Comment).
GB/T 34590-2017 (ISO 26262): Road vehicles — Functional Safety.
GB/T 39263-2020 (ISO 21448): Road vehicles — Safety of the Intended Functionality (SOTIF).
GB/T 40429-2021: Taxonomy and definitions of levels for driving automation.
GB/T 47025-2026: Intelligent and Connected Vehicles — Methods and requirements for simulation testing of automated driving functions.
The commercialization of autonomous driving technology relies on a systematic and standardized testing and verification framework. We focus on providing full-lifecycle testing solutions for autonomous driving systems, covering Functional Testing, Functional Safety Testing, and SOTIF Testing. We support multi-stage verification across MiL, SiL, HiL, ViL, and Real-Vehicle testing to help customers efficiently complete product validation and safety certification.
From early-stage R&D simulation to final real-vehicle validation, combined with data collection and replay technology, PoleLink provides a full-lifecycle testing solution covering functionality, safety, and compliance.
Note: The recommended star ratings are determined by PoleLink based on a comprehensive assessment of testing efficiency, repeatability, effectiveness, safety, and cost.
Horizontal Coverage: Seamlessly connects the entire lifecycle: MiL → SiL → HiL → ViL → Real Vehicle.
Vertical Coverage: Three-dimensional quality control across Functional Testing, Functional Safety Testing, and SOTIF Testing.
Integrated Verification: Functional and safety testing standards are strictly implemented at every stage.
Intelligent Scene Generation: World Models automatically generate diverse test scenarios to optimize coverage.
High-Fidelity Simulation: Technologies like Gaussian Splatting (3DGS) achieve high-precision scene reconstruction to enhance visual realism.
Confidence Validation: A multi-level verification system ensures the reliability of simulation results and the effectiveness of testing.
Flexible Integration: Supports mainstream testing tools (e.g., VECTOR) to provide the most suitable solutions.
Efficient Development: Supported by the CICT system, allowing for automatic model and code generation to boost efficiency.
Years of Deep Expertise: Mature, full-domain, and full-chain technical capabilities in automotive electronics testing.
We ensure product quality through a step-by-step progression, moving from individual units to entire systems, and from simulation to real-vehicle validation.
(1) MiL/SiL Testing (Model/Software-in-the-Loop)
ADAS/AD MiL/SiL Solution Block Diagram
MiL – Rapid Validation in the Algorithm Design Phase:In the model development stage, control algorithm models are embedded in a simulation environment for closed-loop verification. By injecting virtual sensor signals and vehicle model feedback, the correctness of algorithm logic and the rationality of functional design are verified. This allows for rapid iteration and optimization without generating code, making it the earliest and lowest-cost stage to discover issues.
Test Object: Models.
Verification Focus: Algorithm logic correctness and functional design rationality.
Key Value: Earliest detection of design defects; supports rapid iteration and optimization.
SiL – Code Validation in the Software Implementation Phase:Auto-generated or handwritten embedded code is run in PC or cloud environments for closed-loop testing with virtual controlled object models. It verifies the consistency between code and models, software functional correctness, and fault response mechanisms. It supports massive parallel scenario simulation for rapid coverage of huge volumes of test cases.
Test Object: Embedded software code.
Verification Focus: Code-to-model consistency and software fault response.
Key Value: Validates the correctness of code implementation; supports cloud-based batch concurrency for massive scenario coverage.
Real autonomous driving controllers are connected to a test bench. Real-time simulators emulate sensor signals (video, LiDAR point clouds, radar targets) and vehicle dynamics for hardware-level closed-loop verification. This supports fault injection, boundary scenario testing, and Corner Case reproduction, serving as the core method for system-level functional safety verification.
ADAS/AD HiL Solution Block Diagram - Based on Sensor Raw Data Simulation
ADAS/AD HiL Solution based on Sensor Raw Data Simulation
1) Solution Principle:
Test Object: ADAS/AD controllers (driver assistance functions, parking functions, etc.).
HIL Hardware Platform: Includes RT Rack real-time simulators, CAN/CANFD and Vehicle Ethernet communication cards, power management modules, etc.
Simulation Software:
Virtual Scene Simulation Software (e.g., VTD, SimOne): Provides road traffic models, scene rendering, and sensor modeling (including millimeter-wave radar, camera, ultrasonic, and LiDAR).
Vehicle Dynamics Software (e.g., DYNA4): Provides vehicle dynamics models for systems like steering, braking, and powertrain.
Test Management Software (e.g., CANoe): Provides management and execution environments for the HIL system, with configurable I/O resources and fault injection.
Automation Testing Software (e.g., vTESTstudio): Provides automated simulation setup, script execution, and test management.
Sensor Simulation: The system generates various virtual sensor outputs based on the scene simulation software, which are injected into the vehicle controller via interface/protocol adaptation to serve as equivalent replacements for real sensor inputs. Specifics include:
Camera Video Injection: Virtual scene camera outputs are converted into video streams via injection/conversion equipment and injected as raw images into the domain controller's deserializer front-end, replacing the real camera module. (A video dark box solution can be considered for front-view all-in-one camera testing).
Millimeter-Wave Radar Simulation (Target/Point Cloud): Scene software outputs radar model data, supporting both target lists and point cloud simulation, injected via CAN/CANFD or Ethernet to replace the real radar.
Ultrasonic Radar Simulation: Scene software outputs ultrasonic model data to simulate obstacle distance and other information, injected via protocols like DSI3 to replace real ultrasonic sensors.
LiDAR Point Cloud Simulation: Virtual LiDAR models generate point cloud info (distance, intensity, confidence, etc.), sent to the controller via Ethernet encapsulation to replace the real LiDAR.
Integrated Navigation Simulation (GNSS/IMU): Generates positioning and inertial navigation info (latitude/longitude/altitude, attitude, acceleration, etc.), input to the controller via Vehicle Ethernet or CAN bus to replace the real GNSS/IMU.
Map Simulation: Imports OpenDRIVE and other map data into the scene platform to provide the basis for road and traffic elements for scene construction and sensor simulation.
2) Verification Focus: System functionality and performance, hardware fault injection (Functional Safety), and boundary scenario validation (SOTIF).3) Key Value: Automated testing of massive scenarios (scenario generalization) to improve coverage; repeatable verification of dangerous scenarios, accident cases, and Corner Cases; rapid localization of major defects.
ADAS/AD HiL Solution Block Diagram - Based on Sensor Ground Truth (ObjectList) Data Simulation
ADAS/AD HiL Solution based on ObjectList Data Simulation
1) Solution Principle:
This solution uses a Bypass principle for sensor simulation, bypassing the controller's internal perception fusion module. Scene simulation software performs sensor modeling and outputs perceived ground truth data (ObjectList Data), which is injected into the ADAS/AD controller via middleware (e.g., UDP, TCP, SomeIP, DDS, CyberRT, ZMQ, etc.).
Other components (simulation software, HIL hardware platform, etc.) remain consistent with the Raw Data simulation solution.
2) Verification Focus: Functional logic of planning and control algorithms.3) Key Value: Bypasses complex sensor simulation equipment and controller perception modules; lightweight HiL equipment specifically for testing planning/control algorithms.
Note: PoleLink provides these two HiL solutions for system-level verification during controller integration. One supports end-to-end testing of real controllers using Raw Data (Video, LiDAR point clouds, DSI3), while the other provides a lightweight method to test planning/control algorithms using ObjectList data injected via Ethernet.
ViL on Test Beds Solution Block Diagram
ViL on Test Beds:A real vehicle or chassis is placed on an indoor dynamometer. Hub-coupled dynos simulate longitudinal and lateral loads while integrated with virtual scenes for closed-loop testing. This allows for the reproduction of various conditions and extreme scenarios in the lab, providing wider coverage and more precise control, unaffected by weather or site limits.1) Solution Principle:
Test Object: Real autonomous vehicle (virtual sensors, virtual roads).
Virtual Scene & Simulation Environment: Scene software constructs virtual roads and traffic participants. Driver and sensor models (Camera, LiDAR, Radar, Ultrasonic) provide high-fidelity virtual sensor data. Vehicle dynamics models on real-time platforms calculate vehicle responses and road loads to be used as dynamometer inputs.
Autonomous Driving HiL Platform: Sensor simulation equipment injects virtual data into the real vehicle controller. The real-time unit processes virtual data and generates control commands (torque, steering force, wheel speed) sent via EtherCAT to the dynamometer system.
Dynamometer Bench & Control System: The system coordinates hub dynamometers (road resistance/inertia simulation), steering loading motors (steering resistance/lateral load simulation), and pedal loading mechanisms (pedal force feedback) to replicate real-world forces indoors.
Closed-loop Feedback: All actuator states (wheel speed, steering force, pedal position, body attitude, etc.) are fed back to the platform to adjust control commands in real-time.2) Verification Focus: Multi-system coordination, vehicle dynamics response, vehicle-level Functional Safety, and SOTIF.3) Key Value: Repeatable testing of dangerous scenarios under real dynamics.
ViL on Proving Ground Solution Block Diagram
ViL on Proving Ground:In a real proving ground, the real vehicle is overlaid with virtual traffic. Onboard equipment receives virtual targets (vehicles, pedestrians) to verify vehicle-level functions under real road and dynamics conditions, allowing for safe, repeatable testing of dangerous scenarios within the site's physical boundaries.1) Solution Principle:
Test Object: Real vehicle and road (virtual sensors).
Environment: Scene software provides high-fidelity virtual sensor data.
Sensor Simulation: The connection between real cameras and the controller is disconnected, replaced by video injection cards, Ethernet cards, CAN/CANFD cards, and DSI3 cards to inject virtual images and point clouds.
Closed-loop Feedback: Real movement data (attitude, positioning) is synced back to the virtual vehicle in the scene software.2) Verification Focus: Multi-system coordination, vehicle dynamics response, vehicle-level Functional Safety, and SOTIF.3) Key Value: Repeatable testing of dangerous scenarios under real road loads and dynamics.
Note: ViL is joint verification during vehicle integration. PoleLink offers both Test Bed and Proving Ground options. Test Bed ViL uses dynamometers to simulate wheel loads for wider scenario coverage; Proving Ground ViL only simulates sensor data and is limited by the physical site boundaries.
Real Vehicle Testing Solution Block Diagram
Test Object: Real vehicles in real road environments.
Verification Focus: Data collection, vehicle-level functional safety, and high-mileage testing.
Key Value: Final environment validation and regulatory compliance.
Note: Real-vehicle testing is the final verification stage. PoleLink provides data collection and replay equipment for model training and regression testing, as well as onboard fault injection equipment for functional safety testing and overall testing services.
As modular architectures (Perception-Planning-Decision-Control) evolve toward end-to-end (Large Model) solutions, new challenges arise:
Scenario Data: Massive and diverse data is required.
Simulation Fidelity: Environments need higher realism and credibility.
Evaluation: Black-box characteristics mean traditional rule-based evaluation no longer applies.
PoleLink provides a complete data closed-loop solution to address these:
World Model Technology: Generative AI generates diverse scenarios, predicts traffic behavior, identifies key scenes, and ensures scenarios follow physical and traffic rules.
Scene Reconstruction: Uses Gaussian Splatting (3DGS) for extreme visual realism, real-time rendering, and high-fidelity data for various sensors.
Simulation Confidence Verification: A multi-level system (models, data, results) and sensor calibration ensure the simulation accurately reflects reality.
PoleLink has years of experience in automotive electronics testing. Leveraging VECTOR’s leading toolchain, we offer full-lifecycle solutions covering functionality, Functional Safety, and SOTIF. We are committed to technical innovation to ensure the safe and rapid deployment of autonomous driving technology.
© 2022 Shanghai PoleLink Information ICP-Licence:沪ICP备14018904号-1
