HinweisCookies
Diese Webseite verwendet Cookies. Diese dienen der Zwischenspeicherung bei Bestell- oder Anmeldevorgängen. Nicht erfasst werden Daten wie Nutzungshäufigkeit oder Verhaltensweisen. Hier erfahren Sie mehr zum Thema Datenschutz.
The overall goal of the proposed contribution BehAVe is a scenario mining and generation pipeline for the virtual validation and verification of Highly Automated Driving functions (HAD). Whereas model-based simulation lacks a proof regarding the model’s correctness and validity, our data-driven approach promises an explicit validity since the behaviors of the surrounding traffic participants are directly derived from real-life observations of traffic scenarios. In order to reach this goal, we propose an automatic identification of critical scenarios and inference of reasons in terms of behaviors, which contributed most to the outcome of a specific scenario. For further generalization and additional closed-loop testing, we aim to re-generate this scenario constellations in virtual frameworks like CARLA.
Projektakronym | BehAVe | |
---|---|---|
Projektpartner | FZI Forschungszentrum Informatik, Intel Labs Germany (as part of the Intel Collaborative Research Institute (ICRI) - Safe Automated Vehicles | |
Internetpräsenz | http://icri-save.de/ | |
Kontakt | http://icri-save.de/ | |
Beginn | 01.10.2019 | |
Ende | 30.09.2021 | |
Projektlaufzeit | 24 | |
Projektvolumen [Mio. €] | 0,2 | |
Kernthemen | Automatisierung | |
Untersuchungsgegenstände | The current development process of highly automated driving functions (HAD) envisages an accompanying evaluation during development, from components over (sub-)systems to the whole vehicle. As these systems-under-test (SuT) are tested by using different methods, the validity of the test results crucially depends on the applied testing models. In particular in so called X-in-the-loop tests stimulating the SuT on different abstraction levels and interfaces, the validity of the used testing models is inevitable. The validity of the tests is directly linked to the validity of the simulation models for the traffic environment, for example including the optical and material modelling of the geometric surroundings, the environmental effects on sensors as well as the behavior of all traffic participants. Moreover, validity is required especially for the scenarios, under which the SuT is evaluated. In order to pursue such a data-driven approach, a big amount of low-noise and consistent observations from real-life scenarios is necessary. The Test Area Autonomous Driving Baden-Württemberg (TAF-BW) provides the necessary infrastructure to be used as data source for urban scenarios. By utilizing already existing sensors in the infrastructure and the use of high-precision data-processing components, traffic scenarios can be observed over time and will provide valuable training data in terms of spatio- temporal behavior of all traffic participants. | |
Forschungsschwerpunkt 1 | Automatisiertes Fahren,Ableitung verkehrlicher Entscheidungen | |
Forschungsschwerpunkt 2 | Fahrzeugseite,Infrastrukturseite | |
Forschungsziel | Weiterentwicklung | |
Funktionskategorie | Sicherheit,Effizienzsteigernd | |
Verkehrswegkategorie | Stadtstraße | |
Format der Ergebnisse der Rouduser | Results have been published at: https://github.com/fzi-forschungszentrum-informatik/test-area-autonomous-driving-dataset | |
Standort 1 | [49.011196410622105, 8.422919782398864] Testfeld Autonomes Fahren Baden-Württemberg | |
Genutztes Testfeld | Testfeld Autonomes Fahren Baden-Württemberg | |
Karten-Übersicht | ||
Karte 1 | ||
Format | Lanelet2 | |
Genauigkeit Absolut | < 30cm | |
Genauigkeit Relativ | < 3% (< 3cm/m) | |
Auflösung | nicht zutreffend | |
Lizenz | nicht zutreffend | |
Verfügbarkeit | Auf Anfrage. Karten sind Bestandteil des Nutzungsangebots. |