[VIDEO] L4 Autonomy Needs ISO 19237

ISO 19237 was an overly simplistic standard even in 2017, when it was first released for L2 ADAS. It’s goal was to help car makers show they could reduce pedestrian accidents. It defined a single test case, where a pedestrian robot moves directly into the path of the car moving at a constant speed on a straight line. There were a few speed variations for both car and pedestrian robot, and night-time detection was optional. The pass condition was simple: reduce speed to 10 km/h or less at the point of contact, or avoid collision altogether.

Nine years later, it still stands as one of the few normative, behavioral standards for defining safe driving. ISO 21448, in the meantime, a standard intending to address the nuance and uncertainty of the very definition of “safe driving” for ADAS, has had the unintended consequence of applying to L4 autonomy and prolonging and compounding such uncertainty for L4 autonomy. It has very, very limited functionality in the conversation of AV safety requirements, if any at all. In the meantime, however, L2 & 3 ADAS standards have continued to be published, such as the Euro NCAP LKAS, EU Reg 151, and Reg 159. Yet, these industry standards are largely ignored by L4 AV developers and regulators.

Video summary:

• How to assess risk for AVs, even in ISO 19237 situations

• The dangers of ignoring functional safety practices in AVs

• How ISO 19237 compares to naturalistic driving conditions

• Can all AV companies learn from one company’s experience?

• Viewing Waymo’s collision with a child at school from ISO 19237