Predicting Safety Benefits of Automated Emergency Braking at Intersections - Virtual simulations based on real-world accident data
Doctoral thesis, 2018
Introduction: Intersections are a global traffic safety concern. In the United States, around half of all fatal road traffic accidents take place at intersections or were related to them. In the European Union, about one fifth of road traffic fatalities occur at intersections.
Intersection Automated Emergency Braking (AEB) seems to be a promising technology with which to address intersection accidents, as information retrieval by on-board sensing is operational on its own, and, in critical situations, braking is initiated independent of driver reaction. This is not the case for Vehicle-to-Everything (V2X) communication, which requires all conflict-involved vehicles to be equipped with this technology and drivers to respond to an initiated warning. The objective of this thesis is to evaluate the effectiveness of a theoretical Intersection AEB system in avoiding accidents and mitigating injuries. As it will take several decades for a new safety technology to penetrate the vehicle fleet and full coverage of all vehicles may never be achieved, the technology benefit is here analyzed as a function of market penetration. Finally, this research assesses whether a set of test scenarios can be derived without compromising the variance of real-world accidents.
Methods: Data from the United States National Automotive Sampling System / General Estimates System and the Fatality Analysis Reporting System was used to compare the capacity of on-board sensing and V2X communication to save lives. To investigate Intersection AEB in detail, the German In-Depth Accident Study (GIDAS) data and the related Pre-Crash Matrix (PCM) were utilized to re-simulate accidents with and without Intersection AEB using different parameter settings of technical aspects and driver comfort boundaries. Machine learning techniques were used to identify opportunities for data clustering.
Result: On-board sensing has a substantially higher capability to save lives than V2X communication during the period before full market penetration of both is reached. The analysis of GIDAS and PCM data indicate that about two thirds of left-turn across path accidents with oncoming traffic (LTAP/OD) and about 80 percent of straight crossing path (SCP) accidents can be avoid by an idealized Intersection AEB. Moderate to fatal injuries could be avoided to an even higher extent. Key parameters impacting effectiveness are vehicle speed and potential path choice; to increase effectiveness, these should be limited and narrowed down, respectively.
Conclusion and Limitations: Intersection AEB is effective in reducing LTAP/OD and SCP accidents and mitigating injuries However, intersection accidents are highly diverse and accurate performance evaluation requires taking variations into account. The simulations were conducted using ideal sensing without processing delays and an ideal coefficient of friction estimation.
straight crossing path
left turn across path