Reliably identifying relevant and safety-critical traffic scenarios is a central challenge in developing automated driving systems (ADSs). The development of objective measures to detect the intuitively perceived level of criticality across various scenarios remains unresolved. The major challenge lies in the evaluation of all sensitive metrics of all participants in a scenario, which goes beyond assessing individual criticality metrics or providing only non-operational definitions. To address these challenges, a modular and practice-oriented framework that assesses traffic criticality is proposed. Vehicular and environmental data can be used to objectively measure individual critical elements of traffic, which are then aggregated into a comprehensive scenario assessment. The framework is illustrated by applying it (i) to two maneuvers discussing effects of single criticality metrics and (ii) to two widely used datasets featuring highway and intersecting traffic for a statistical analysis. The assessment framework can be applied in many scenarios, ranging from identifying safety critical scenarios in existing data as a fundament for early development of ADSs to detecting safety critical emergency maneuvers in live data to trigger a deployed ADS. Fundamental to this process is the suitable selection of metrics, thresholds for critical behavior, and the system’s sensitivity in aggregating partially critical scenarios.


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    Assessing Critical Traffic Scenarios: A Modular Evaluation Framework


    Beteiligte:


    Erscheinungsdatum :

    01.06.2025


    Format / Umfang :

    2140936 byte




    Medientyp :

    Aufsatz (Zeitschrift)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch





    SIMULATING TRAFFIC SCENARIOS

    CHEHADE ABDALLAH / KASSAR SARI / SAVARGAONKAR MAYURESH VIJAY et al. | Europäisches Patentamt | 2025

    Freier Zugriff

    Scenario Factory: Creating Safety-Critical Traffic Scenarios for Automated Vehicles

    Klischat, Moritz / Liu, Edmond Irani / Holtke, Fabian et al. | IEEE | 2020


    Simulation of Traffic Scenarios

    Kirmse, A. / VDI Gesellschaft Fahrzeug- und Verkehrstechnik | British Library Conference Proceedings | 2003


    Flow: A Modular Learning Framework for Mixed Autonomy Traffic

    Wu, Cathy / Kreidieh, Aboudy / Parvate, Kanaad et al. | ArXiv | 2017

    Freier Zugriff