What a software-defined vehicle actually is
The term is used loosely enough to mean almost nothing. Here is the concrete engineering definition, what genuinely changes, and how to tell a real SDV claim from marketing.
7 min read
SDV
Cars are becoming computers with wheels — fewer, more powerful processors, functions delivered as software, features that arrive after the car is sold. This is what that actually means in engineering terms, covering both the instrument cluster and the infotainment domain, with no assumed vocabulary.
What a software-defined vehicle actually is, the pressures driving the change, and how vehicle electronics got here.
The term is used loosely enough to mean almost nothing. Here is the concrete engineering definition, what genuinely changes, and how to tell a real SDV claim from marketing.
7 min read
SDV is expensive, disruptive and organisationally painful. Four pressures explain why car makers are doing it anyway — and knowing them makes most architecture decisions predictable.
5 min read
Distributed, then domain, then zonal, then central — four architectures, what each solved, what each broke, and how to tell which one a programme is actually on.
6 min read
Where the industry actually is versus where the slides say it is — what has genuinely shipped, what has gone wrong, and how to read an SDV claim without being cynical about it.
5 min read
The roles that appear, the ones that change, and the skills that carry across — written for someone deciding what to learn next rather than for a strategy deck.
5 min read
Zonal wiring, central compute, hypervisors and mixed criticality — the hardware shape that makes software-defined possible.
The hardware shape that makes software-defined possible — why grouping by location beats grouping by function, and what moves to the centre when it does.
6 min read
Several operating systems on one chip, with proof that one cannot disturb another — what a hypervisor actually does, and why it is the enabling technology for consolidation.
6 min read
The idea that makes features addable after production — components offering named capabilities others discover and call, and the interface discipline it demands.
6 min read
Packaging vehicle software so it can be updated independently — what containers buy you in a car, what they cost, and where they are genuinely inappropriate.
5 min read
The move from broadcast signals to callable services: SOME/IP, DDS, Automotive Ethernet, VSS and the data broker.
The physical layer that makes everything else possible — why a car needed its own Ethernet, and why raw speed was never the hard part.
5 min read
The protocol that carries vehicle services — discovery, methods, events and fields — and the design decisions that determine whether an interface survives a decade.
5 min read
The other major vehicle protocol — publish-subscribe with quality-of-service guarantees. Why ADAS chose it, why body electronics did not, and how to pick between it and SOME/IP.
5 min read
A standard naming scheme for vehicle data, and why the boring problem it solves — three teams disagreeing about what a signal means — is the one that actually costs programmes months.
6 min read
The component that holds current vehicle values and serves them to everyone — decoupling producers from consumers, and making the whole stack testable without a vehicle.
5 min read
AUTOSAR Classic and Adaptive, Android Automotive, Linux and AGL, QNX — what each is for and how they coexist.
The two AUTOSAR platforms, why they exist, what each is actually for, and why a modern vehicle runs both at once rather than migrating from one to the other.
9 min read
An honest assessment of what AAOS contributes to a software-defined vehicle, what it deliberately does not do, and where the boundary between Android and the rest of the vehicle actually sits.
8 min read
The open-source alternative to Android in the cockpit — how an automotive Linux is actually built, what Automotive Grade Linux provides, and the real trade-off an OEM is making when it chooses this route.
7 min read
Why a commercial microkernel keeps winning the safety-critical slot, what certification actually buys, and how a safety OS sits next to Android on the same chip.
8 min read
Safety-rated display, rendering strategies, and what happens when the cluster and infotainment share one chip.
The screen behind the wheel is the one with legal obligations. How it is architected when everything else is consolidating, and what Android is and is not allowed to own.
6 min read
Four ways to put content on the cluster — native, streamed, shared-surface and hardware-composed — with the trade-offs that decide which one a programme picks.
6 min read
What has to be proven before a consolidated cockpit can ship — the evidence a safety case actually needs, and the requirements that keep landing on the wrong system.
6 min read
IVI as a consumer of vehicle services, app platforms, the cockpit domain controller, and the store model.
The screen with the most freedom and the least authority — what changes for IVI when the vehicle becomes service-oriented, and why that is mostly good news.
5 min read
Writing infotainment software against vehicle services rather than raw signals — discovery, absence, version skew, and rendering honestly when the vehicle disagrees with you.
6 min read
How software gets onto a vehicle and stays updatable — system image, OEM store, public store or sideload — and what each route costs in release cadence and capability.
5 min read
Virtual ECUs, continuous integration for vehicles, digital twins, OTA and feature-on-demand.
Running vehicle software without vehicle hardware — what a virtual ECU actually is, the four levels of fidelity, and why this is the change that makes everything else in an SDV possible.
8 min read
What a real automotive CI pipeline looks like, why the last mile is nothing like web deployment, and how safety, regulation and physical hardware reshape a familiar practice.
8 min read
What a digital twin is once the marketing is removed, the four kinds that actually get built, and how to tell a useful one from an expensive demonstration.
8 min read
How software actually reaches a sold vehicle — the difference between updating a head unit and updating a car, the preconditions nobody mentions, and what goes wrong.
8 min read
Selling software capability after the vehicle is built — how entitlement actually works end to end, why some attempts were rejected by customers, and the engineering it forces on you.
7 min read
What a fleet produces, where it is processed, what may leave the vehicle, and who is accountable for it.
What a fleet actually produces, why almost none of it can be sent anywhere, and the filtering decisions that determine whether a data programme is useful or merely expensive.
7 min read
Deciding what runs on the vehicle and what runs in a datacentre — the four constraints that make the decision for you, and the hybrid patterns that come out of it.
6 min read
A vehicle knows where you go, who is inside and how you drive. What the law requires, what good practice looks like, and the specific decisions that land on a cockpit engineer.
8 min read
Functional safety across a consolidated stack, zero trust inside the vehicle, and regulation that now has teeth.
ISO 26262 in plain terms, why consolidation makes safety harder rather than easier, and what changes when the software is supposed to be updated after the safety case was written.
8 min read
Why the vehicle network stopped being a trusted place, what replaces implicit trust between components, and the practical controls that follow.
7 min read
R155 and R156 turned security process into a condition of selling vehicles. What a CSMS actually requires, what an SBOM is for, and how a vulnerability report becomes a fleet-wide fix.
7 min read
One feature built the whole way through — from a physical sensor to both the cluster and the centre screen.
A complete worked example — a seat massage feature built from the physical actuator to both the cluster and the centre screen, naming every layer, every interface and every decision along the way.
7 min read