Some of the new products on show
DAY 2: Waymo explores computational metrics that can be used as behavioral benchmarks
Waymo
Azadeh Dinparastdjadid, senior research scientist at Waymo, is in San Jose this year to discuss automated driving system behavior evaluation in the 'Regulations, standards, metrics and collaboration – building blocks for safe ADAS and AD technologies' session.
The presentation covered how computational metrics can be used as behavioral benchmarks for flagging when ADS behavior does not meet expectations. Dinparastdjadid also explained how such metrics can also be applied to discovering novel scenarios and events with behavioral issues from existing datasets.
Speaking to AAVI, she highlighted a particular concept that Waymo uses: “Drivership is the realization of expected driving behaviors that position the ADS as a good citizen of the road, which can build trust by meeting and exceeding societal expectations on how autonomous driving systems behave on the road. Behavior reference models are a means to operationalize Drivership. These benchmarks can include any metric, or criterion used to operationalize proper/improper Drivership. Waymo’s models such as NIEON, Surprise and FSM will help operationalize Drivership.”
Read more in her recent interview with AAVI here.