[Question]: Automated testing #127
Description
System Information / Logs
As you know, with the support of AI, production efficiency has improved significantly. Therefore, testing and verification are now the main concerns for autonomous driving.
How to AI-driven then?
Question
As you know, with the support of AI, production efficiency has improved significantly. Therefore, testing and verification are now the main concerns for autonomous driving.
1. Automate Everything
Eliminate manual steps to scale verification.
- Auto mine high-value scenarios.
- Trigger full CI/CT on code commits.
- Use synthetic data for rare corner cases.
2. Focus on Processes, Not Rules
Evaluate through structured workflows, not static checklists.
- Compare baseline vs candidate vs human in shadow mode.
- Introduce adversarial (“red-team”) stress testing.
- Enforce a deterministic pipeline from code → SIL → HIL → closed course → open road.
3. Evaluate Multi-Dimensionally and Reproducibly
Ensure metrics are quantifiable and repeatable.
- Track safety, comfort, and efficiency.
- Guarantee deterministic replay of tests.
- Auto generate KPI comparisons and block regressions.
AI-driven
What I Tried
No response
Relevant Logs or Links
No response