How do regulators test a self driving system before approving public road use

Reviewing Extensive Simulated and Closed-Course Testing Data

Regulators generally begin by reviewing extensive simulated and closed-course testing data submitted by the autonomous vehicle developer, evaluating how the system performed across a wide range of scenarios tested in a controlled environment before the vehicle is ever considered for actual public road testing.

Requiring Supervised On-Road Testing With a Safety Driver

Before granting broader approval, regulators generally require a period of supervised on-road testing with a trained human safety driver present, capable of taking control if the autonomous system encounters a situation it can’t handle safely, providing real-world validation beyond what simulated and closed-course testing alone can demonstrate.

Evaluating Documented Performance Across Rare Edge Cases

Regulators also evaluate the system’s documented performance specifically across rarer, more unusual edge-case scenarios, since a system performing well under typical, common driving conditions doesn’t necessarily guarantee reliable performance during the rarer situations that, while infrequent, still carry genuine real-world safety consequences if handled poorly.

Why Specific Requirements Vary Meaningfully by Jurisdiction

Specific testing requirements and overall regulatory rigor vary meaningfully across different jurisdictions, since autonomous vehicle regulation, particularly within the United States, is largely handled at the state level rather than through a single unified federal standard, meaning a system approved for testing in one state doesn’t automatically carry approval elsewhere.

Why This Fragmented Regulatory Landscape Creates Genuine Complexity

This fragmented regulatory landscape creates genuine complexity for autonomous vehicle developers seeking broader deployment, since expanding operation into a new jurisdiction often requires navigating that jurisdiction’s own specific testing and approval requirements, rather than a single approval automatically applying nationwide or globally.

Bottom Line

Regulators test self-driving systems by reviewing simulated and closed-course data, requiring supervised on-road testing with a safety driver, and evaluating performance across rare edge cases, though specific requirements vary meaningfully across jurisdictions given the largely state-by-state nature of autonomous vehicle regulation in the U.S.

Go deeper

• How do regulators test insurance ai models for unfair discrimination before approval?
• What is a sandbox program and how do regulators use it to test ai rules before finalizing them?
• How is safety testing for physical robots different from testing software only ai?