Mathematical safety proofs for ADVs
National Institute of Informatics

Researchers from the National Institute of Informatics (NII) in Tokyo, Japan, will present their technique for establishing mathematical and rigorous safety proofs for automated driving vehicles (ADVs). The technique – Goal-Aware RSS – relies on the RSS methodology (Responsibility-Sensitive Safety, Shalev-Shwartz et al. 2017), but features logical formalization and dedicated software for writing complex safety proofs.

With Goal-Aware RSS, users can derive a mathematical safety rule compliance with which guarantees safe completion of a designated driving scenario, even when the driving scenario is complex (such as an emergency pull over on a highway). Those safety rules can be retrofitted to any existing ADV controller as a safeguard mechanism where the safety rules intervene and execute safety maneuvers when needed.

The safety rules derived using Goal-Aware RSS can be thought of as mathematical driving laws. They can thus play many important roles in the ADV ecosystem to come – in safety assurance, international standards, laws and regulations, insurance and so on. In particular, those safety rules can be a key enabler for the wide deployment of ADVs in regions such as Europe and Japan, serving as a mechanism to ensure and explain user level safety.

Booth: 2046

Back to News