White Paper

Data Annotation And Autonomous Vehicle Safety

The increased attention to Autonomous Vehicle safety issues over the past 2-3 years reflects the evolution in car and system development – and a bit of irony in public perception. Car manufacturers were quick to position early industry successes in AV Level 1 and Level 2 as victories for enhanced driver safety. But public perception has shifted as car manufacturers move on to Level 3 and Level 4 system development, where the aim shifts to complete autonomy of the vehicle. It was inevitable that the underlying focus on the part of the developers would expand from improving the basic functional capabilities to operating safely within complex on-road scenarios. That has been reflected in subtle but clear changes in data analysis and annotation priorities.

Download the White Paper to:

  • Understand the evolving focus on autonomous vehicle safety and its correlation to public perception
  • The changes in data annotation workflows, as increased on-roading testing takes place
  • Dive into the emergence and nuances of safety standards like ISO 26262 and UL 4600
  • Learn how the Uber crash of 2018 impacted the development journey of AVs and its regulatory consequences

First 300 words

How safe is safe enough? When the comparison is between a car driven by a computer versus a person, the practical answer depends on competing perspectives. 

Purely from a statistical standpoint, Autonomous Vehicle developers have a clear target. The United States alone sets a grim annual benchmark. According to the National Highway Traffic Safety Administration, there are more than six million car crashes every year. More than 36,000 Americans die in those crashes, with another 2.5 million ending up in hospital emergency rooms. 

Considered globally, the figures are even more sobering.

What if a switch to fully Autonomous Vehicles cut those figures in half, could AV developers declare victory? Perhaps from the standpoint of cold logic, but those involved in various AV initiatives would concede that the more telling dynamic is consumer confidence – as evidenced by their resulting willingness to buy a fully autonomous vehicle, or even step inside one. Developers grimace at the thought of having to defend the idea of tens of thousands of deaths from collisions involving driverless cars.

A 2018 Rand Corporation report, “Measuring Automated Vehicle Safety,” alludes to the conflict between the need for empirical data for AV developers seeking to advance the state of the art, and the determination of consumer safety regulatory bodies to resist what they generally regard as avoidable risk.

“In the United States – and elsewhere, to some degree – the emergence of AVs has been associated at least implicitly with the view that some exposure to risk and uncertainty about this risk must be accepted in the short and medium terms to see the long-term benefit of AVs,” the report notes. “(But) consultations for this project showed how that view , held by AV developers and by associated industrial and research partners, is not held by the…