RESUMEN
Objective: This study aimed to assess the effectiveness of autonomous emergency braking (AEB) systems in car-to-cyclist frontal collisions by simulating their effects, in terms of crash avoidance and injury mitigation, on a representative target population of real-world accidents. Identifying effectiveness-critical AEB-cyclist design parameters through a sensitivity analysis was also targeted.Methods: The analysis is based on a representative set of real-world car-to-cyclist frontal collisions gathered from French police reports. AEB-cyclist-relevant accident cases were first selected and used to build injury risk curves for fatal, severe, and slight cyclist injuries. The effect of AEB-cyclist on these cases was then simulated by means of a car kinematic model involving sensor detection strategies and actuator actions. Combining the resulting simulated impact speed distributions with the injury risk curves allowed to assess AEB-cyclist's effectiveness in terms of lives saved and mitigated injuries. Using design of experiments methods, the sensitivity of this effectiveness with regards to AEB-cyclist design parameters could be assessed.Results: Cyclist injury risks curves were built, along with their confidence intervals, for fatal, severe, and slight injuries using a polytomous complementary log-log regression model, with squared impact speed as an independent variable. A sensitivity analysis on an ideal bisensor AEB-cyclist setting highlighted influential design parameters such as maximal braking intensity or crucial decision algorithm parameters such as maximal time and distance to collision thresholds. AEB-cyclist effectiveness was nevertheless shown to range from 35% to 59% in fatalities, 14% to 54% in severe injuries, and 11% to 42% in slight injuries, depending on field of view parameters alone, once reference values of decision algorithm parameters had been set.Conclusions: This study illustrates the potential benefits and limits of AEB-cyclist systems. High-end systems show acceptable effectiveness rates, but road safety performance strongly depends on external factors such as road surface conditions or has to be tuned in order to avoid unnecessary activations and driver discomfort. Limits of the system's everyday use (lack of maintenance, driver reaction time to collision warnings, etc.) were not taken into account, thus resulting in optimistic evaluations of AEB-cyclist effectiveness.
Asunto(s)
Accidentes de Tránsito , Automóviles , Ciclismo , Desaceleración , Algoritmos , Recolección de Datos , Urgencias Médicas , Francia , Humanos , Policia , Estudios Prospectivos , Equipos de Seguridad , Heridas y LesionesRESUMEN
Objective: With the overall goal to harmonize prospective effectiveness assessment of active safety systems, the specific objective of this study is to identify and evaluate sources of variation in virtual precrash simulations and to suggest topics for harmonization resulting in increased comparability and thus trustworthiness of virtual simulation-based prospective effectiveness assessment. Methods: A round-robin assessment of the effectiveness of advanced driver assistance systems was performed using an array of state-of-the-art virtual simulation tools on a set of standard test cases. The results were analyzed to examine reasons for deviations in order to identify and assess aspects that need to be harmonized and standardized. Deviations between results calculated by independent engineering teams using their own tools should be minimized if the research question is precisely formulated regarding input data, models, and postprocessing steps. Results: Two groups of sources of variations were identified; one group (mostly related to the implementation of the system under test) can be eliminated by using a more accurately formulated research question, whereas the other group highlights further harmonization needs because it addresses specific differences in simulation tool setups. Time-to-collision calculations, vehicle dynamics, especially braking behavior, and hit-point position specification were found to be the main sources of variation. Conclusions: The study identified variations that can arise from the use of different simulation setups in assessment of the effectiveness of active safety systems. The research presented is a first of its kind and provides significant input to the overall goal of harmonization by identifying specific items for standardization. Future activities aim at further specification of methods for prospective assessments of the effectiveness of active safety, which will enhance comparability and trustworthiness in this kind of studies and thus contribute to increased traffic safety.