Page 32 - EE Times Europe Magazine - June 2025

32 EE|Times EUROPE

OPINION | AUTONOMOUS DRIVING | SAFETY AND SECURITY

True SDVs Are Out of OUTDATED REGULATIONS RISK CREATING
‘SOFTWARE-DEFINED LIABILITIES’
Reach in Automotive’s The SDV promise is real: richer functionality,
improved safety, and a step change in user
experience. But current development and
Legacy Model regulatory models weren’t built for this world.
The way we develop and validate auto-
motive software is fundamentally broken.
The existing approach can’t keep pace with
By John Ellis, Codethink the complexity and scale of modern vehicle
systems. It is rooted in slow-moving safety
Even the best engineers can’t out-design outdated standards, costly certification processes, and
outdated assumptions about software deter-
business structures. Strategic changes are needed to minism. Without a new approach, SDVs risk
becoming expensive, fragile, and impossible
unlock the potential of the software-defined vehicle. to assure at scale.
The ISO 26262 automotive safety standard
is a case in point. Written more than
From autonomy and connectivity to personalized in-vehicle 15 years ago, the standard was relevant to
experiences, the automotive industry is chasing the vision of the small, tightly controlled codebases running
software-defined vehicle (SDV). Yet only a few OEMs have deliv- on simple microcontrollers. ISO 26262 was
ered on the promise of a car that improves meaningfully after it crafted based on the assumption that soft-
leaves the factory. ware was predictable and deterministic—an
The reason? It’s not a lack of engineering talent. The bottleneck assumption that clearly no longer holds.
is structural: entrenched business models and regulatory frame- Today’s vehicles run on hundreds of
works that constrain even the brightest teams. Winning the SDV millions of lines of code, executed across
race will require confronting both head-on. multicore microprocessors with deep memory
hierarchies, L1/L2 caches, and speculative
LEGACY BUSINESS STRUCTURE BLOCKS SDV PROGRESS execution—one of many techniques silicon
Bringing a true SDV to life demands bold, structural changes, starting with software owner- manufacturers use to enhance execution
ship. Without such ownership, you’re just stitching together parts from others and never truly speed and throughput. This modern
seeing the whole system. hardware/software stack introduces inherent
Tesla’s edge, for example, comes not just from innovation but from integration and owner- nondeterminism, meaning that the same
ship. Nearly all of its software is built, owned, and maintained by Tesla employees. In contrast, software can behave differently under differ-
most legacy OEMs rely heavily on Tier 1 and Tier 2 suppliers. That structural dependency ent execution conditions. The standards that
fragments both responsibility and capability. currently govern safety-critical automotive
Conway’s Law applies here: Organizations build systems and products that mirror their software never accounted for this reality.
own structure. Siloed teams produce siloed software. The prevailing “Yellow Book” develop-
ment style—where engineering is outsourced and abstracted away from the problem—only A NEW GOAL: TRUSTABILITY
reinforces this disconnect. Until OEMs restructure governance and remove friction from It’s time to move past the illusion that
engineering, progress will stall. exhaustive testing and traditional design
Consider something as basic as the model year. In the legacy automotive world, everything ensure safety. That supposition no longer
hinges on model years. OEMs hold new features and technologies for the next model year scales. The real question is: How do we trust
release, hoping to incentivize buyers to purchase new models. Production schedules and complex, nondeterministic systems?
inventory management are built around the model-year milestone. On the regulatory side, Winning OEMs will redefine their North
model years are used to track compliance with standards and safety. Their use is embedded in Star. Instead of chasing compliance alone,
both the business model and the regulation of the industry itself. they will focus on trustability, building audit-
But in a true SDV world, software evolves continuously. Features can be purchased, down- able, verifiable processes that account for the
loaded, and activated at any time. The SDV model supports ongoing sales and deeper customer realities of modern software.
engagement, but legacy thinking blocks its path. This means rethinking governance, restruc-
turing for software ownership, and aligning
SMALL STRUCTURAL CHANGES UNLOCK BIG IMPACT business and regulatory models with contin-
One simple but impactful structural change that can enable engineering is a shift in how uous delivery. That’s the path to true SDVs,
revenue is recognized. Indeed, look to accounting principles to open the door for business and the companies that follow it will lead the
models defined by software. industry. ■
The iPhone serves as an example. Under a traditional accounting model, Apple would
record the full sale price at purchase. But by deferring some revenue and recognizing it over John Ellis is the president and head of
time—tied to software updates and support—Apple aligns accounting with continuous value product at Codethink, a provider of critical,
delivery. high-performance software projects for
This shift matters. If a car is truly a “data center on wheels,” its business model should international-scale companies in industries
reflect that. Continuous software delivery demands continuous support, maintenance, and including automotive, finance, medical,
monetization. That means rethinking not just engineering but finance. and IoT.

JUNE 2025 | www.eetimes.eu