Software-defined vehicles: Architecture, benefits, and industry impact

Design a vehicle once, manufacture it, and sell it as-is for the next five years while working on the next one. Thought of a new feature that its drivers or passengers would appreciate in the meantime? Too bad: you can’t add it without racking your brain over compatibility issues with dozens of other electronic control units (ECUs) or physically tinkering with the hardware.

But what if you could add new features just as easily as Apple pushes iOS updates to iPhone users?

That’s the promise of software-defined vehicles. Instead of dozens of function-specific ECUs, SDVs consolidate compute into a smaller number of high-performance controllers, while simpler ECUs remain at the edge.

The result? Vehicles become “smartphones on wheels”: easily updatable and customizable.

In a traditional vehicle, each function is defined by hardware, limiting long-term flexibility. In a software-defined vehicle (SDV), in turn, functions are governed by software. That makes updating these functions and adding new ones times easier than in hardware-defined cars.

This is the definition that seems to draw consensus in the automotive industry. However, other definitions exist. For instance, PwC defines SDV as an ecosystem that “provides new experiences or value to the user” through both in-car and out-of-car functions.

While touchscreens and infotainment systems aren’t exactly new to vehicles, they don’t make those vehicles software-defined. A software-defined car is built using a drastically different architectural approach. Instead of a hundred separate ECUs, it uses only a few, with software taking control of functions such as driving assistance, connectivity, and so on.

Software-defined vehicles are also different from connected and autonomous vehicles:

• Connected vehicles are vehicles with internet access and vehicle-to-everything (V2X) communication capabilities. While SDVs also have similar connectivity capabilities, unlike connected vehicles, they rely on a different internal architecture to enable upgrading core functions through over-the-air (OTA) updates.

• Autonomous vehicles are vehicles capable of driving themselves without human input under certain conditions. Not all SDVs come with autonomous driving capabilities, and not all autonomous vehicles have to be software-defined.

Here are the hardware and software layers that make up a software-defined vehicle architecture:

Hardware

In a traditional vehicle, many functions are handled by separate ECUs. That requires a lot of extra wiring and connections, which makes cabling crowded and, at times, redundant.

In a software-defined vehicle, on the other hand, the hardware is connected based on its location (zone). In other words, SDVs increasingly adopt zonal or hybrid architectures that reduce hardware coupling and enable centralized compute.

The kind of hardware present in a software-defined vehicle is also different. Zonal controllers aggregate signals locally, while centralized Systems on a Chip (SoCs) execute high-level vehicle functions. An SoC is akin to a mini-computer; it’s equipped with the computing and memory capabilities needed to perform complex computational tasks. For example, Intel offers SoCs for SDVs.

Certain software capabilities (think AI computations for autonomous driving) also require adding hardware accelerators to the software-defined vehicle architecture: CPUs, GPUs, or a combination of the two.

Software stack

In a software-defined vehicle platform, the software stack comprises several layers:

• Operating system(s) manage the computing and memory capabilities of the hardware

• Middleware enables communication between the hardware and applications and maintains critical functionality (diagnostics, security, logging)

• Basic software and runtime services enable basic operations and connectivity

• Functional software enables core functionality (ADAS, perception, planning)

• Application software powers functions directly visible to the user (e.g., infotainment, navigation, driving assistance)

Notably, in software-defined vehicles, hardware and software are decoupled to enable seamless over-the-air updates. The middleware and operating system make this decoupling possible. They serve as abstraction layers, providing a standardized interface for user-facing automotive industry software solutions.

Cloud integration

Cloud platforms complement SDVs by enabling development, analytics, fleet operations, and OTA pipelines, while core vehicle functions remain edge executed. The cloud can partially host application data or perform back-end operations remotely to power certain capabilities.

For example, software-defined vehicles can use cloud integration for:

• Data analytics

• Data storage and management

• CRM integration

• Vehicle simulation and validation

• Partner ecosystem applications

By shifting the focus from hardware-enabled functionality to software-enabled performance, SDVs can evolve long after they leave the production line. However, that’s just one of the many software-defined vehicles benefits.

Over-the-air (OTA) updates

One overlooked flaw in the component design could lead to a large-scale vehicle recall, which may total millions or even billions in repair, replacement, administrative, and legal costs. While SDVs may still fall victim to faults such as door-latching issues, reliance on software makes it possible to address specific issues through bug-fixing updates.

The result? In many cases, OTA updates can reduce or avoid recalls related to software defects.

Over-the-air (OTA) updates don’t require customers to visit repair shops; they’re delivered as easily as a smartphone app update. Besides bug fixing, these updates can:

• Enhance the performance of existing functionality

• Add new capabilities, as long as the hardware supports them

• Maintain the software’s security

For example, Tesla added pothole detection in a software update in 2022.

OTA updates can do more than provide just bug fixes and quality-of-life tweaks. They can also include enhanced safety updates, such as better emergency braking or collision avoidance. These features, based on real-time data and analytics, are also more effective at preventing accidents, thus improving road safety for everyone.

For SDVs equipped with advanced driver-assistance systems (ADAS), reliance on real-time data and analytics also improves decision-making, making it faster in critical situations. 

Operational efficiency and cost reduction

In a software-defined car, variables such as engine parameters and battery pack performance can be continuously tracked, analyzed, and optimized. For instance, software can monitor fuel consumption or engine performance and adapt to maximize efficiency. Or, if the vehicle is electric, software can independently optimize the charging cycles to extend battery life.

Predictive maintenance, in turn, can reduce maintenance costs and prolong the vehicle’s lifespan. By collecting and analyzing component data in real time, predictive maintenance solutions can detect issues before they cause a breakdown and prompt the driver to visit the nearest repair shop if necessary.

New business models for OEMs

Traditional vehicles are bought once, generating no additional revenue for the manufacturer until the customer decides to purchase a new model. But as SDVs can seamlessly gain new features through OTA updates, automakers can sell new add-on functions to existing customers throughout the car’s life cycle.

For example, Ford made its hands-off driving feature available for an extra $75 per month in 2023, while Tesla sells its Autopilot for $99 per month.

As appealing as this idea is, it remains to be seen whether consumers will be ready to open their wallets and pay for add-on subscriptions. For example, when BMW tried to charge $18 a month for seat warmers in 2022, the backlash was so severe that the company abandoned the idea in 2023.

Software-defined vehicles enable many use cases throughout their life cycle, with the most noteworthy being:

• Design and engineering: Shorter feature development and deployment cycle to experiment with new features and functions

• Production: Vehicle self-testing to speed up validation

• Maintenance and service: Predictive maintenance to prevent breakdowns, recall prevention with bug fixes, vehicle performance logging for faster and more effective repairs

• Consumer personalization: Infotainment personalization, driving experience customization with profiles for multiple drivers

• Value-added services and monetization: Partner ecosystem of user-facing applications, data sharing and monetization, on-demand paid upgrades

• Next-gen telematics: Proactive maintenance without technician involvement, use case-specific functions tailored to customer segments

• Fleet management: Data sharing with rental car companies or fleet managers

The shift toward software-defined vehicles has two major implications for the automotive industry:

• SDVs shake up the established revenue model, adding new recurring monetization opportunities

• Due to reliance on software, SDVs require a more substantial investment into both initial software development and continuous maintenance

To adapt to this shift, automakers need to move away from a waterfall methodology to an agile one. A centralized electrical and electronics architecture will, in turn, have to replace distributed embedded software development.

This isn’t the only business implication of the shift to SDVs. To remain competitive, automakers also need to:

• Transition to the life cycle economics to optimize the costs (e.g., software updates) and revenues (optional on-demand features) of SDV development

• Focus on the customer instead of product tech to provide an outstanding customer experience both inside and outside the car

Adopt a data-driven approach to development that prioritizes applying the data collected from customer vehicles and virtual development and validation processes

The shift to a software-first architecture is challenging, and so is the decoupling of software and hardware. In fact, 79% of auto executives cite the latter as a moderate or significant SDV challenge, due to:

• Software complexity. More code equals a higher risk of bugs and vulnerabilities going unnoticed, while integrating multiple systems and layers can easily cause a headache.

• Cybersecurity risks. Software can be hacked, taken over, or exploited for access to sensitive data. When the software in question powers a vehicle, a cyberattack may immobilize it or hijack its controls.

• Data privacy concerns. SDVs collect vast amounts of data about the vehicle, its performance, and user behavior to function. That raises concerns over how this data is stored, processed, and used, especially if it’s shared with third parties.

• Potential consumer pushback. As mentioned above, backlash against paid add-ons like BMW’s seat-warming subscription can be brutal. That backlash can undermine the trust toward the brand and cost automakers customer loyalty.

• Regulatory compliance. Compliance with existing data privacy regulations already represents a challenge. In addition, changing vehicle behavior via OTA updates poses legal questions regarding compliance with safety regulations and accident liability.

• Lack of standardization. As it stands, the SDV market lacks standardization across platforms, operating systems, and cloud environments. This can cause compatibility, scalability, and interoperability issues in the long run.

• OTA update risks. While they’re convenient, pushing safety-critical and infrastructure updates must be done carefully to avoid compatibility issues or system failures.

• Talent gaps. Most OEMs don’t have a vast engineering team on the payroll, especially when it comes to cutting-edge functionality like ADAS features. Building the workforce with the expertise required for SDV development is a process that may take years.

Is software the future of automaking? Three-quarters of auto industry executives believe so, expecting the software-defined experience to become the core brand value by 2035.

Today, software-defined vehicles already represent more than the highly personalized, continuously evolving future of mobility. They are the foundation for autonomous vehicles (AV) and electric vehicles (EV) alike, helping the former evolve faster and enabling smarter energy management for the latter (see Figure 5. The relationship between SDVs and autonomous driving/electrification development, What is an SDV [Software Defined Vehicle]? Defining SDVs beyond just vehicles, pwc, 10-03-2024).

AI-defined vehicles are emerging as a possible next leap in the evolution of SDVs. These vehicles will rely on AI systems in every aspect of their functioning, from infotainment personalization to predictive collision avoidance. The software for AI-defined vehicles, however, will require even more computational power for resource-intensive edge computing.

The automotive industry is on its way to SDV-as-the-new-normal, and OEMs have to adapt to this shift as the industry is gradually moving toward more collaborative development models.