4,000 tests every night.

About the customer

This customer is a Tier 1 supplier of connected car systems, including infotainment and audio, to automakers worldwide. The name is withheld under the programme's confidentiality terms.

Each programme ships across many vehicle configurations and models, and every one of them has to be validated on real hardware.

The challenge

Every infotainment programme ships across many vehicle configurations and models. For the supplier's test teams that meant:

• Slow adaptation. Adapting test scripts to new configurations and models took long.
• Manual fallback. Scripts broke often, so a lot of the testing went back to people on the bench.
• Unstable tests. Frequent hardware and state changes, and the bugs that come with them, made results hard to trust.
• Long release cycles. Testing set the pace of every release.

What the team does now

• 4,000 tests every night. The regression runs unattended overnight on the benches.
• In the CI/CD pipeline. Runs are triggered from the pipeline, and the team gets a report every day.
• Infotainment and audio. Both are covered by the same suite.
• The bench is part of the test. A step sends a brake signal through a CAN tool; the next step checks on the infotainment screen that the brake lights of the car model are lit.

How a CAN check reads

A test step can call a bench tool as easily as it taps the screen:

1. Send the brake signal with the CAN tool.
2. Open the vehicle view on the head unit.
3. Verify that the brake lights of the car model are lit.

The tool is wired in once, as a custom tool, and every step, screen or bus, lands in the same report.

The result

The nightly suite replaced the manual bench runs for the covered tests. Configuration changes no longer mean rewriting scripts, and the daily report shows the state of every build.