A software-defined vehicle (SDV) is a car whose features are set mainly by software running on a small number of powerful computers, and which keeps gaining or changing features through over-the-air updates after it is sold. Designing one is hard because it combines the scale of a large software project with the safety, security, cost and lifetime demands of automotive hardware. The ten challenges below cover software development, cybersecurity, system integration, OTA updates, AI, hardware compatibility, regulation, user experience, data privacy and cost, followed by the role semiconductors play in making SDVs work.
The automotive industry is going through a major shift with the arrival of software-defined vehicles. Traditional vehicles are defined largely by their hardware; SDVs put software at the centre of what the vehicle can do. The change is driven by advances in connectivity, artificial intelligence and data processing, which let vehicles become more adaptive, more intelligent and better integrated with digital services.
SDVs promise a range of benefits, from better safety features and automated driving to personalised driving experiences and seamless over-the-air updates. Getting there is not simple. This post looks at the top 10 challenge areas in designing and deploying SDVs and at the approaches being used to address them, for industry engineers and interested readers alike.
Key takeaways
- SDVs move from dozens of single-function ECUs towards a few central and zonal computers, which makes integration easier in the long run and much harder during the transition.
- Software scale is the root of most challenges: a modern car is commonly reported to carry on the order of 100 million lines of code or more.
- Safety (ISO 26262, ISO 21448), cybersecurity (ISO/SAE 21434, UN R155) and software update rules (UN R156) apply together and must be designed in from the start.
- OTA updates are the defining SDV capability and need secure boot, signed images, rollback and careful power management.
- Semiconductors set the limits: central compute SoCs, automotive Ethernet, secure elements, and wide-bandgap SiC and GaN power devices in electric vehicles.
The 10 challenges at a glance
| # | Challenge | Main issues |
|---|---|---|
| 1 | Software development | Managing massive codebases, agile methodologies, ensuring code quality |
| 2 | Cybersecurity | Protecting against vulnerabilities, implementing encryption, regular updates |
| 3 | System integration | Integrating ECUs, seamless communication, testing for interoperability |
| 4 | OTA updates | Reliable mechanisms, patch management, minimal disruption |
| 5 | AI integration | Autonomous driving, safety compliance, continuous training |
| 6 | Hardware compatibility | Software-hardware co-design, performance optimisation, hardware constraints |
| 7 | Regulatory compliance | Adhering to standards, keeping up with regulations, documentation requirements |
| 8 | User experience | Intuitive interfaces, user testing, aesthetic design |
| 9 | Data privacy | Data protection measures, complying with laws, transparency practices |
| 10 | Cost management | Balancing costs, resource allocation, cost-effective solutions |
| Semiconductors to the rescue | Wide-bandgap (WBG) semiconductors: silicon carbide (SiC) and gallium nitride (GaN) |
1. Software development
- Managing massive codebases, often reported to run to a hundred million lines or more across a vehicle, written by the carmaker and many suppliers.
- Implementing agile methods with continuous integration and deployment, which has to coexist with the V-model processes that safety standards and Automotive SPICE expect.
- Ensuring code quality through rigorous testing, code reviews, static analysis and coding standards such as MISRA C and AUTOSAR C++14 guidelines.
In practice this means separating hardware-dependent code from applications, using middleware such as AUTOSAR Classic and Adaptive, and running large automated test farms with software-in-the-loop and hardware-in-the-loop rigs.
2. Cybersecurity
- Protecting against vulnerabilities in vehicle software, from telematics and infotainment to the brake and steering controllers behind them.
- Implementing encryption and secure communication protocols, for example authenticated in-vehicle messages (AUTOSAR SecOC), TLS to the cloud and hardware security modules in each critical ECU.
- Updating security measures regularly to counter new threats, with monitoring and incident response for the life of the vehicle.
ISO/SAE 21434 defines the engineering process, and UNECE Regulation No. 155 requires a certified cybersecurity management system for type approval in the markets that apply it.
3. System integration
- Integrating electronic control units (ECUs) that control vehicle functions. Older vehicles can have around a hundred separate ECUs; SDV architectures consolidate them into domain or zonal controllers and a central computer.
- Ensuring seamless communication between subsystems over CAN, CAN FD, LIN and automotive Ethernet, often with time-sensitive networking for deterministic traffic.
- Testing for interoperability and performance across many scenarios, vehicle variants and software versions.
4. Over-the-air (OTA) updates
- Designing reliable and secure update mechanisms: signed images, secure boot and a chain of trust from the cloud to each ECU.
- Handling patch management and rollback, typically with A/B (dual bank) flash so a failed update can return to the previous version.
- Keeping disruption to the vehicle to a minimum, for example by downloading in the background and installing only when parked with enough battery charge.
UNECE Regulation No. 156 sets requirements for a software update management system, including records of which software is in which vehicle.
5. Artificial intelligence
- Integrating AI for functions like automated driving, driver monitoring and predictive maintenance.
- Ensuring AI systems comply with safety standards. ISO 26262 was not written for machine learning, so ISO 21448 (safety of the intended functionality, SOTIF) and ISO/PAS 8800 for AI in road vehicles are used alongside it.
- Continuously training models with real-world data, then validating and redeploying them through the OTA pipeline.
6. Hardware compatibility
- Ensuring new software features work on hardware that was fixed when the car was built, which means planning spare compute, memory and bandwidth for years of updates.
- Performing hardware-software co-design to optimise performance, for example mapping neural networks onto the accelerators in the central SoC.
- Testing for hardware constraints and limits such as thermal throttling, memory bandwidth and worst-case latency.
7. Regulatory compliance
- Adhering to standards such as ISO 26262 for functional safety, plus ISO 21448, ISO/SAE 21434 and Automotive SPICE.
- Keeping up with evolving regulations and industry standards, which differ between regions.
- Ensuring documentation and processes meet regulatory requirements, including traceability from requirement to test for every software release.
8. User experience
- Designing intuitive and user-friendly interfaces for vehicle interaction, including voice and touch, without distracting the driver.
- Conducting user testing to gather feedback and improve design.
- Balancing aesthetic design with functional and safety requirements.
9. Data privacy
- Implementing data protection measures to safeguard user information such as location history, driver images and voice recordings.
- Complying with data privacy laws and regulations such as the EU GDPR and national data rules in other markets.
- Being transparent about data use and collection, with clear consent options for the owner.
10. Cost management
- Balancing the costs of software development, testing and deployment against the bill of materials.
- Optimising resource allocation to maximise efficiency.
- Seeking cost-effective solutions without compromising quality, for example by reusing one compute platform across several models.
Illustrative example: consolidating many small ECUs into a few zonal controllers and one central computer adds cost in the computer but can remove wiring and connectors, which are a significant share of vehicle electrical cost and weight. Whether the total falls depends on the vehicle and volume, so carmakers model it case by case.
Semiconductors to the rescue
Semiconductors, especially wide-bandgap (WBG) semiconductors such as silicon carbide (SiC) and gallium nitride (GaN), play a crucial role in SDVs, particularly electric ones. They offer superior electrical and thermal properties, enabling higher efficiency, faster charging and greater durability in electric vehicles (EVs). WBG semiconductors can handle higher voltages, temperatures and switching frequencies, which makes them well suited to power electronics, electric drivetrains and charging systems.
| Chip type | Role in an SDV |
|---|---|
| Central compute and ADAS SoCs | Run most vehicle software and AI models; need spare headroom for future updates |
| Zonal and safety microcontrollers | Real-time control with lockstep cores for high ASIL levels |
| Automotive Ethernet PHYs and switches | High-bandwidth backbone between zones and the central computer |
| Hardware security modules and secure elements | Key storage, secure boot and signature checks for OTA |
| SiC MOSFETs | Traction inverters and high-voltage (for example 800 V) systems |
| GaN transistors | Compact on-board chargers and DC-DC converters |
FAQ
What is the difference between a software-defined vehicle and a connected car?
A connected car has network access, for example for navigation or remote apps. An SDV goes further: its core functions are defined in software on central computers and can be changed through updates over its life.
Which standards matter most for SDV software?
ISO 26262 for functional safety, ISO 21448 for SOTIF, ISO/SAE 21434 for cybersecurity and UNECE R155 and R156 for cybersecurity and software update management, with Automotive SPICE for process.
Why are SiC and GaN linked to SDVs?
Most SDV platforms are also electric. SiC and GaN improve the efficiency of the power electronics, which extends range and speeds up charging, while the software layer manages them.
Your comments will be moderated before it appears here.