Development of Safety Performance Assessment Metrics for ADS-Equipped Vehicles 2020-01-1206
The safety performance of an automated driving system (ADS)-equipped vehicle must be quantified in order to be able to compare to human-driven vehicles and begin to answer the question of "How safe is safe enough for ADS?" In this research, metrics and systems for the measurement and analysis of said metrics are defined and developed that capture aspects of the safety of ADS-equipped vehicles as they operate on public roadways, interacting with both human-driven and ADS-equipped vehicles.
The metrics that quantitatively describe the safety of an ADS-equipped vehicle from both on- and off-board data sources are first defined, including those that are focused on collisions, safety envelope violations, and rules-of-the-road violations. Each identified metric has a data range as well as its data source (including both on- and off-board sources), within the context of an operational design domain (ODD). These metrics allow the measurement of the sequencing involved in crashes, dangerous situations and near misses, enabling the evaluation of their temporal flows and identifying their key aspects that lead to actual impacts and dictate the severity of said impacts. Second, an intersection in Anthem, AZ is instrumented with a sensor suite appropriate for the measurement of the metrics associated with off-board data sources. The research includes determination of sensing needs and implementation of the designed systems. Third, artificial intelligence- and computer vision-based algorithms are developed to apply the metrics to the collected data. The objective is to determine what data source (or combination of data sources) is/are required to effectively measure and analyze the metrics that will allow for a safety performance assessment to be made of an ADS-equipped vehicle.
Jeffrey Wishart, Steven Como, Maria Elli, Brendan Russo, Jack Weast, Niraj Altekar, Emmanuel James
Exponent Inc., Arizona State University, Intel, Northern Arizona University, University of Arizona