Propelmee, a self driving technology start-up, discusses safety assurance and insurability of self driving cars in its latest safety publication CARSAV by proposing novel perspectives on validation of self driving technologies, drawing from techniques and methods from insurance telematics for autonomous vehicles.
Comprehensive Architecture for Road Safety of Autonomous Vehicles (CARSAV) presents novel thinking on a range of topics related to verification and validation of highly and fully automated driving systems and unpacks the complexity of safety assurance requirements for commercial deployment of autonomous vehicles. The publication explains how autonomous vehicle technologies work, how they come together to enable automated driving, what are the key aspects in these technologies that are safety critical and how to address safety gaps.
“Our goal in publishing this report is to provide an argued position on the current technological approaches with respect to their safety dimension and how to bridge the gaps,” said Zain Khawaja, CEO and founder, Propelmee.
“To that end, we put forth the argument that in order to achieve Comprehensive Autonomous Safety(CAS), every single technology component comprising the full autonomous system must be verifiable and be able to conform to formal validation at a component level in order to establish comprehensive system-level safety.”
CARSAV introduces the idea of Classification Agnostic Perception (CAP) as the first guarantor of safety in automated driving systems and expounds on the importance of a system design approach that enables transparent scrutiny and assessment. “We emphasize the importance of assessing and validating each component technology in the self driving software system as well as validating the tools used for verification and validation of such systems. We draw upon insurance telematics techniques to propose incident analysis methods for fault assignment in automated driving systems. Unless the system is fully transparent and can prove its performance on that basis, it will not be possible to earn approval for commercial deployment from regulators and trust from end users.”
For more information, visit here.
