Reference Design
Automotive Control System
A complete reference design for building production-ready automotive systems with ADAS integration, functional safety compliance, and comprehensive vehicle networking capabilities.
Introduction
What This Reference Design Is About
This automotive reference design provides a complete blueprint for developing Advanced Driver Assistance Systems (ADAS) and automotive control units. It demonstrates our proven approach to building safety-critical automotive electronics that meet rigorous industry standards while remaining adaptable to various vehicle platforms and applications.
The design covers everything from sensor fusion and object detection to vehicle control and networking—enabling Level 2+ autonomous driving capabilities with a clear path to higher autonomy levels.
Requirements
- Functional Safety: ISO 26262 ASIL-D compliance with redundant processing
- Real-Time Performance: Deterministic processing with <10ms latency
- Environmental: -40°C to +125°C operation, vibration resistance
- Connectivity: CAN/CAN-FD, Automotive Ethernet, V2X communication
- Security: Secure boot, OTA updates with rollback protection
Recommended Architecture
Distributed processing architecture with clear separation of safety-critical and non-critical functions:
- • Primary ECU: High-performance SoC for AI/ML processing and sensor fusion
- • Safety MCU: Lockstep ARM Cortex-R5F for safety monitoring and fail-safe control
- • Sensor Hub: Dedicated processing for camera, radar, and LiDAR fusion
- • Gateway Module: Secure vehicle network communication and firewall
- • Power Management: Intelligent power distribution with redundancy
Essential Parts of This Design
- • Multi-camera vision processing (up to 8 cameras, 4K resolution)
- • Radar integration (77GHz automotive radar modules)
- • LiDAR sensor interface and point cloud processing
- • Object detection and classification using deep neural networks
- • Path planning and trajectory optimization algorithms
- • Vehicle control interface with fail-safe mechanisms
- • V2X communication for cooperative driving features
- • OBD-II and UDS diagnostic protocols
This reference design serves as a customizable foundation—adapt it to your specific vehicle platform, sensor configuration, and autonomy requirements.
Hardware
Recommended Platform: Arches
Arches Platform (NVIDIA Jetson Xavier NX/AGX Orin) is our recommended platform for automotive ADAS applications.
Why Arches:
- • Superior neural network inference for real-time object detection
- • Multi-camera ISP for simultaneous video processing
- • Hardware video encoders for data logging
- • NVIDIA DRIVE ecosystem compatibility
- • Automotive-qualified variants available
Key Specifications: 384-core Volta GPU, 6-core Carmel CPU, 8GB LPDDR4x, hardware video codecs
Essential Components to Add
Components added to baseline Arches platform:
- • Safety MCU: TI TMS570 lockstep processor for ASIL-D monitoring
- • CAN Transceivers: NXP TJA1145 automotive CAN-FD with partial networking
- • Automotive Ethernet: Marvell 88Q5072 1000BASE-T1 switch
- • GMSL Deserializers: Maxim MAX96712 quad GMSL2 for camera input
- • Radar Interface: SPI/CAN interface for automotive radar modules
- • Power Management: Automotive PMIC with 7-36V input, watchdog
- • Secure Element: NXP SE050 for key storage and secure boot
Schematic Capture & PCB Layout
Design Tools: Altium Designer with automotive component libraries
PCB Stack-up: 12-layer HDI with controlled impedance for high-speed signals
Key Considerations:
- • Separated power domains for safety isolation
- • GMSL and Ethernet differential pair routing
- • Thermal management with embedded copper planes
- • EMC considerations for automotive compliance
Regulatory Requirements
Functional Safety:
- • ISO 26262 ASIL-D for safety-critical functions
- • Hardware Fault Metrics (PMHF, LFM) compliance
Environmental:
- • AEC-Q100 qualified components
- • ISO 16750-3 vibration and shock
- • IP67 ingress protection
EMC: CISPR 25 Class 5, ISO 11452-2
RTOS/Operating System
Recommended OS Configuration
Hybrid Architecture: Linux + RTOS running on separate cores
Primary SoC (Jetson):
- • Yocto Linux with PREEMPT_RT patches
- • Handles AI inference, sensor processing, HMI
- • Non-safety-critical path planning
Safety MCU (TMS570):
- • SafeRTOS (ASIL-D certified RTOS)
- • Safety monitoring and vehicle control
- • Fail-safe actuation commands
Why This Configuration
Linux for AI/ML: Required for NVIDIA TensorRT, CUDA, and deep learning frameworks. Rich ecosystem for computer vision and sensor processing.
RTOS for Safety: Deterministic timing guarantees required for ISO 26262 compliance. Certified RTOS provides documented safety evidence.
Separation of Concerns: Safety-critical functions isolated from complex software, simplifying safety certification.
Interprocessor Communication
Communication Channels:
- • Dedicated CAN bus between SoC and Safety MCU
- • SPI for high-bandwidth data transfer
- • GPIO for interrupt signaling and heartbeat
Safety Protocol: CRC-protected messages with sequence counters and watchdog monitoring for communication integrity.
Kernel & Driver Configuration
Kernel Configuration:
- • PREEMPT_RT for real-time scheduling
- • CPU isolation for critical threads
- • High-resolution timers enabled
Custom Drivers Developed:
- • GMSL camera drivers with V4L2 integration
- • CAN-FD driver with error injection support
- • Automotive Ethernet AVB/TSN driver
- • Radar SPI interface driver
Middleware
Recommended Middleware: DDS
eProsima Fast DDS - Data Distribution Service for real-time automotive data communication.
Why DDS for Automotive:
- • Quality of Service (QoS) policies for deterministic delivery
- • Peer-to-peer communication without central broker
- • Scalable publish-subscribe for sensor data
- • AUTOSAR Adaptive Platform compatible
Regulatory Considerations
AUTOSAR Compatibility:
- • AUTOSAR Adaptive ara::com API mapping
- • Service-oriented architecture support
Cybersecurity:
- • ISO/SAE 21434 cybersecurity compliance
- • DDS Security plugin for encrypted communication
- • Access control and authentication
Middleware Interfaces
Sensor Data Topics:
- • Camera image streams (H.265/RAW)
- • Radar object lists and point clouds
- • LiDAR point cloud data
- • IMU/GPS fusion data
Vehicle Interface:
- • CAN/CAN-FD message abstraction
- • Vehicle state and dynamics
- • Actuation commands
- • Diagnostic services
Application Interface:
- • Perception output (objects, lanes)
- • Planning trajectories
- • System health and diagnostics
- • Configuration management
Application
Application Architecture
Embedded Linux Application with real-time processing pipeline for ADAS functions.
Architecture Pattern: Component-based with DDS communication, allowing independent deployment and testing of each module.
Design Decision: C++17 for performance-critical components, Python for prototyping and testing tools.
Core Application Modules
- • Perception: Object detection, classification, and tracking using TensorRT-optimized neural networks
- • Sensor Fusion: Multi-sensor fusion algorithms for robust environmental perception
- • Localization: GPS/IMU/Visual odometry fusion for precise positioning
- • Planning: Path planning and trajectory optimization with safety constraints
- • Control: Vehicle control algorithms with fail-safe mechanisms
Development Tools
Essential Tools:
- • NVIDIA TensorRT for neural network optimization
- • NVIDIA DeepStream for video analytics pipeline
- • OpenCV for computer vision algorithms
- • Eigen for linear algebra computations
Testing & Simulation:
- • CARLA simulator for scenario testing
- • Vector CANoe for vehicle network simulation
- • Google Test for unit testing
Diagnostics & Security
Log Management & Debugging:
- • Structured logging with DLT (Diagnostic Log and Trace)
- • Event data recorder for incident analysis
- • Remote debugging via secure shell
OTA & Security:
- • Secure boot chain with TPM
- • A/B partition scheme for safe updates
- • Signed firmware with rollback protection
Other Essential Supporting Apps
Cloud Integration
- • Fleet Management: Vehicle tracking, status monitoring, and remote diagnostics
- • OTA Update Server: Secure firmware distribution with delta updates
- • Data Analytics: Driving data collection and analysis for model improvement
- • Map Services: HD map updates and crowd-sourced map corrections
Desktop Applications
- • Calibration Tool: Camera/radar alignment and parameter tuning
- • Diagnostic Dashboard: Real-time visualization of sensor data and system health
- • Data Labeling: Tools for annotating driving data for model training
- • Log Analyzer: Post-processing tools for incident analysis
Web-Based Apps
- • Fleet Dashboard: Web portal for fleet monitoring and management
- • Technician Portal: Mobile-friendly diagnostics and service tools
- • Report Generator: Automated safety and compliance reporting
- • Training Platform: ML model training and validation dashboards
Our Experience
Products Delivered
Using this reference design, we have successfully delivered:
- Commercial Vehicle ADAS: Level 2 autonomous system for delivery trucks with collision avoidance and lane keeping
- Agricultural Autonomy: Autonomous navigation system for agricultural vehicles with GPS-denied operation capability
- Mining Vehicle Safety: Proximity detection and collision avoidance system for underground mining vehicles
Key Achievements
50%
Faster Time to Market
ASIL-D
Safety Certification
10ms
End-to-End Latency
8
Camera Support
Our automotive reference design has been validated in real-world deployments across multiple vehicle platforms and operating conditions.