Software-defined and AI-defined vehicles

Software‑defined vehicles need software‑defined validation.

Test your ECU software from the first line of code to the car on the road.

Book a demo

01  /  The products

From a drafted test project to a real ECU on the benchA Studio window drafts a test project and runs it against a simulated ECU. Silicon arrives: the ECU becomes real, a bench unit is wired to it with a fault injected on one wire and current measured on another, and the coverage report fills.TEST_PROJECTSTUDIOplant model — from the descriptionAI ASSISTEDtest script — from the specAI ASSISTEDscenario — built and replayedtarget — vECU, before hardware existstarget — ECU, real siliconCOVERAGE REPORTvECU · SIMULATEDECU · REAL SILICONVALIDATIVE UNITCIAFAULT INJECTEDCURRENT MEASURED

One application to write and run the tests.See Studio

One bench for when the code has to meet real silicon.See the hardware

02  /  The platform

Write the test once. Run it everywhere.

One test project runs in simulation, on Validative hardware, and on benches already in use.

Author once in Validative Studio

The desktop application where tests are written. One test project in version control, running on every commit.

Run before silicon exists

Production code against a simulated vehicle, on a normal PC.

Run the same tests on real hardware

Real flashed binary, real bus traffic, fault injection.

Virtual ECUCAN & LINFault injectionScenario sweepsCI pipelines
threshold_suite5 testsIllustrative example
Input above the threshold raises the alarmPass0.42 s
Input just below the threshold stays silentPass0.38 s
Input exactly at the threshold triggersPass0.40 s
Absolute limit reached, the unit enters the safe statePass0.91 s
Sensor open circuit during a rising inputSkipped—
Same test file. Target changed, not the tests.Simulated vehicle

Connect Studio to existing HIL rigs with an adapter, built for each rig. dSPACE, Vector, NI, Typhoon and rigs built in-house.

Studio runs on Windows. Tests are Python files in a repository.

See how Studio works

03  /  Works with existing hardware

Keep the bench. Lose the queue.

Every hardware target sits behind the same interface. A rig already in use becomes one more place the tests can run.

emergency_braking_suite211 scenariosIllustrative00:00
Queued
Software in the loopOn every commit
feat/braking-limitspushed 2 min ago
Scenarios executed0 / 211
Host runner, no hardware—
Only if the step above passed
Waiting
Hardware in the loopNightly
Fault injectionsensor signal lost
Scenarios executed0 / 211
Real ECU, flashed binary—
Same test suite.Different cadence, different target.
One test projectversion controlled
Adapter layer
Simulated vehiclehost runner
Validative hardwarereal ECU on the desk
Existing rigalready in use

The same suite, on a desk unit or on the rig already in the lab.

See the hardware
emergency_braking_suiteIllustrative
208 of 211 covered3 gaps

04  /  AI-defined vehicles

AI-defined functions need AI-scale validation.

A function that learns has no single right answer to check. Proof comes from covering the space, not from counting passes.

100% passed is not the same as 100% tested.

BoundaryNothing tested just either side of the speed where the function stops working.
NegativeSituations where it must not brake were never run.
EquivalenceOne situation was left to stand for a whole group.

Switch to pass rate → the same run, told the flattering way

Every run is reported the same way, whichever target produced it.

See how coverage is reported

See it run

You’re building the future of mobility.Don’t let the road become your test bench.

Twelve minutes. A real control function, a seeded bug, and a pipeline that catches it before anyone touches hardware. No slides.

The pilot package is scoped to one ECU. Tell us the unit, the interfaces it uses and the tests you run today — four fields, no call required.

The pages are not live yet. An engineer can answer in the meantime.