HMI TestingMay 1, 2025
Why Scripted Tests Fail Modern Automotive Software
As vehicles become software-defined platforms, traditional test automation can't keep up. Here's why — and what comes next.
Automotive cockpit & HMI software
Every release, proven.
Filuta validates cockpit and HMI software exhaustively within a declared scope and delivers the evidence for every release — the way UN R156 now requires. In production at a European automotive group: nine years of manual tester-time now runs in ten days per release.
Every software release is now a compliance event
UN R156 requires an approved Software Update Management System — version traceability and demonstrated safe execution — for every software update to a type-approved vehicle.
European Union
July 2024
Mandatory for all new vehicles
Great Britain
1 June 2026
Complete and base vehicles · SI 2025/1110
China
1 January 2026
GB 44495 / GB 44496
Three major markets. Every release. Sampling a fraction of the interface and signing for the rest is no longer a defensible position.
Standards we produce evidence against
UN R155UN R156ISO/SAE 21434ISO 24089ISO 26262ISO/PAS 8800:2024
Filuta produces validation evidence. It is not a certification or type-approval body.
From sampling to proof
Modern cockpit software has more reachable states than any test team can visit. Filuta’s neurosymbolic engine learns a model of the interface and explores it exhaustively within a declared scope — across head units, languages and configurations — and hands back the evidence.
What manual testing cannot give you
Manual and scripted testing sample the interface. Dozens of languages, several head-unit configurations and thousands of menu paths mean the sample is small, the result is not reproducible, and there is no coverage figure to put in a homologation file. The release ships anyway — with someone’s signature on it.
With Filuta:
Declare the scope, get a coverage number
Screens, states, languages and head-unit configurations are declared up front. The engine explores everything inside that scope and reports how much of it was reached.
Reproduce every finding
Every defect comes with a trace that replays it. Engineering fixes what was found; nobody argues about whether it happened.
Re-validate every release, not just the big one
Each software update gets the same exhaustive pass and the same evidence pack — the cadence UN R156 assumes and manual testing cannot keep.
No source code, no instrumentation
The engine works from the interface a driver sees. Nothing has to change in the system under test, and nothing of yours leaves your environment.
Release Assurance
Validation as a subscription, not a project.
Scoped per model line and release train
The scope is declared up front — screens, states, languages, configurations — and the coverage claim is made against it.
Every release in the window, validated exhaustively
Not a one-off project. Each software release inside the contracted window is explored to the declared scope before it ships.
Evidence pack in R156-ready form
A coverage number, a reproducible trace for every finding and a statement of what was proven — delivered per release, in the shape a homologation file needs.
Priced on guaranteed coverage, not tester days
You buy the coverage of the declared scope. How many hours it takes us is our problem.
Tell us the model line and the release train. We come back with the declared scope, the coverage guarantee and the evidence format.
How Can We Help
Validation that produces evidence, not just findings. The engine understands the interface the way a driver does — and explores it the way no driver could.
Autonomous exploration to a declared scope
Define what must hold — 'every navigation path completes', 'media playback works from every input source' — and the scope it must hold in. The engine plans its own way through the interface and checks every state it reaches against the specification.
Cross-configuration consistency
The same specification is checked across head units, screen sizes, languages and vehicle configurations — without a tester visiting each combination by hand.
Visual regression detection
Unintended visual changes, layout shifts and rendering defects across displays are caught release to release, with the before-and-after trace attached.
In production at a European automotive group
Validation of a next-generation infotainment system across 38 languages and multiple vehicle platforms. Nine years of manual tester-time now runs in ten days per release, with no source code and no instrumentation of the target.
Why the oracle must be independent
The problem with self-graded tests
The International AI Safety Report 2026 records that it has become more common for models to distinguish between test settings and real-world deployment, and to exploit loopholes in evaluations. An evaluation whose oracle is the system under test proves nothing.
What Filuta does differently
Our oracle is symbolic and independent of the software it examines, and the scope is declared before exploration starts — which is what turns “we found problems” into a coverage claim.
We also ship no production software. No supplier relationship, no incentive to grade our own homework, and evidence a homologation file can actually use.
Case Studies
What was in scope, what was found, and what was proven — from the cockpit-software deployment at a European automotive group.
For testing and engineering partners
If you already own the customer relationship — a group testing house, an engineering-services provider, a TIC body — Filuta can be the engine inside your validation product. You keep the customer, the contract and the delivery. We supply the exploration engine and the evidence pipeline.
Prove the next release before you sign it.
Tell us the model line and the release train. We come back with the declared scope, the coverage guarantee and the evidence format.
