Driving Safety Performance Metrics Development and Validation 

Abstract
The driving safety performance of automated driving system (ADS)-equipped vehicles (AVs) must be quantified in order to be able to assess the performance and compare it against that of human-driven vehicles. The Institute of Automated Mobility (IAM), created by Executive Order in Arizona to advance safety, policy, and technology of the AV industry, has initiated research to develop an operational safety assessment (OSA) methodology for AVs [1]. The first step in this development is to determine a comprehensive set of metrics to enable an OSA of AVs and human-driven vehicles in a variety of scenarios. This metric set [2] was determined by conducting a broad literature review of existing metrics that have been proposed to assess the driving safety performance of both human-driven vehicles and AVs as well as novel metrics proposed by the IAM research team.

The OSA metrics are outlined as well as the methods being used for validation, which include both simulation in CARLA and real-world testing at the IAM testbed in Anthem, AZ. The validation has initially focused on safety envelope metrics, which are used to indicate whether the subject vehicle is in a safe or unsafe proximity to safety-relevant objects in its vicinity. The safety envelope metrics include metrics developed by the traffic engineering community (e.g., time-to-collision (TTC), modified time-to-collision (MTTC), time headway (THW), and post-encroachment time (PET)) as well as safety envelope metrics based on the Responsibility-Sensitive Safety (RSS) framework developed by Intel/Mobileye [3]. The OSA metrics and OSA Methodology form an integral component of the efforts by the Verification and Validation (V&V) Task Force under SAE International's On-Road Automated Driving (ORAD) Committee to develop standards to help ensure the safe deployment of AVs on public roads.