Abstract:As driverless automated driving systems (ADS) start to operate on public roads, there is an urgent need to understand how safely these systems are managing real-world traffic conditions. With data from the California Public Utilities Commission (CPUC) becoming available for Transportation Network Companies (TNCs) operating in California with and without human drivers, there is an initial basis for comparing ADS and human driving safety.

What problem does this paper attempt to address?

The paper primarily discusses the safety issues of Automated Driving Systems (ADS) on public roads. As more and more autonomous vehicles hit the road, it becomes crucial to understand how these systems handle real traffic conditions and assess their safety. The study is based on data provided by the California Public Utilities Commission (CPUC) and the California Department of Motor Vehicles (DMV), and analyzes the safety of three driving types: Uber's ride-sharing trips, supervised Autonomous Vehicles (AV), and Waymo and Cruise's driverless ADS pilot and deployment projects. All driving takes place within the city of San Francisco, excluding highways, to control for external factors such as vulnerable road users' exposure. The study found that the supervised AV had a similar accident rate per million miles (CPMM) compared to Uber's human-driven vehicles, while Waymo's driverless AV had a lower CPMM, and Cruise's driverless AV had a higher CPMM than Uber's human-driven vehicles. However, the current volume of data is not sufficient to support conclusions about whether currently automated systems are safer or less safe than human-operated vehicles in complex urban environments. The paper also reviewed existing literature, highlighting some gaps in directly comparing automated driving to human-driving accidents, and discussed the limitations and handling methods of the data. The authors analyzed the mileage and accident reports of autonomous vehicles using data from the DMV, CPUC, and other sources in an attempt to bridge these gaps. Through analysis, the study identified accident characteristics of supervised AVs, such as mostly minor damages, around 16% accidents resulting in injuries, and the most common collision type being rear-end collisions. Additionally, the impact of external factors such as weather, lighting conditions, and road conditions on accidents was considered. Despite this, due to limitations in the data, the study could not comprehensively assess the safety of autonomous driving systems, especially concerning vulnerable road users.

Initial Indications of Safety of Driverless Automated Driving Systems

Comparative safety performance of autonomous- and human drivers: A real-world case study of the Waymo Driver

A Survey of Autonomous Driving: Common Practices and Emerging Technologies

Benchmarks for Retrospective Automated Driving System Crash Rate Analysis Using Police-Reported Crash Data

Dynamic Benchmarks: Spatial and Temporal Alignment for ADS Performance Evaluation

Safety Oriented State Transitions in Level 3 Automated Driving Systems: A General Framework.

Is it Safe to Drive? An Overview of Factors, Challenges, and Datasets for Driveability Assessment in Autonomous Driving

Is it Safe to Drive? An Overview of Factors, Metrics, and Datasets for Driveability Assessment in Autonomous Driving

Operational safety hazard identification methodology for automated driving systems fleets

Exploratory analysis of injury severity under different levels of driving automation (SAE Level 2-5) using multi-source data

Comparative Safety Assessment of Automated Driving Strategies at Highway Merges in Mixed Traffic

A matched case-control analysis of autonomous vs human-driven vehicle accidents

Analysis of Advanced Driver-Assistance Systems for Safe and Comfortable Driving of Motor Vehicles

Safe Driving Capacity of Autonomous Vehicles

Driver perceptions of advanced driver assistance systems and safety

Introduction to the Special Section on Connected and Autonomous Vehicles: Technologies, Methodologies, and Applications

Testing of Autonomous Driving Systems: Where Are We and Where Should We Go?

Characterizing Behavioral Differences and Adaptations of Automated Vehicles and Human Drivers at Unsignalized Intersections: Insights from Waymo and Lyft Open Datasets

Development and classification of autonomous vehicle's ambiguous driving scenario

A Data-Informed Analysis of Scalable Supervision for Safety in Autonomous Vehicle Fleets