Claims that self-driving cars could avoid 90% of road deaths by eliminating drivers’ errors are unproven, and shared responsibility between robot and human drivers may actually increase risks.

The report says self-driving cars should be designed so safety-critical systems are functionally independent and cannot fail in case of connectivity issues

That’s the finding of a new report by the International Transport Forum that examines the impact of automated vehicles on road safety and cyber-security vulnerabilities.

The report sets out how shared responsibility between robot and human drivers can lead to more complex driving decisions, which might make driving less safe, not more, including for drivers who don’t normally take risks.

The report also sets out that humans retain an advantage over partially automated systems in many areas but fully automated and connected cars are vulnerable to cybersecurity risks.

“The avoidance of crashes should never depend on access to shared external communication channels alone,” the report warns.

It also recommends designing automated vehicles so that safety-critical systems are functionally independent and cannot fail in case of connectivity issues.

The report also advocates a ‘Safe System’ approach, which organises all elements of road traffic in a way that when one safety mechanism fails, another steps in to prevent a crash, or at least serious injury. Applying this approach to automated driving means the traffic system will account for machine errors.

Another recommendation is that the industry should avoid safety performance being used to market competing automated vehicles.

“Regulators and industry should work together to ensure uniform safety performance of automated driving systems. The relative safety level of vehicles deployed, or strategies employed, should not be a competition issue. The regulatory framework should ensure maximum achievable road safety, guaranteed by industry, as a precondition of allowing these vehicles and services to operate,” it explains.

Other recommendations include:

• Require automated vehicles to report safety-relevant data.
• Develop and use a staged testing regime for automated vehicles.
• Establish comprehensive cybersecurity principles for automated driving.
• Provide clear and targeted messaging of vehicle capabilities.

To view the report, click here.