To content
Fakultät für Informatik
Simulation-based Testing

Representativity of Traffic Scenarios in Verification and Validation of Automated Vehicles

© Falk Howar​/​TU Dortmund

The goal of the project is to develop novel algorithms for assessing the representativity of traffic scenarios from simulations or real-world test drives. These algorithms shall support the V&V of automated vehicles by identifying untested traffic situations and comparing tests with each other.

The goal of this project is to provide a metric that allows estimating the coverage of an Operational Design Domain (ODD). The ODD defines possible entities, e.g., types of vehicles, pedestrians, and road properties, that are relevant for the investigated domain. These entities and relations represent the base for defining abstract predicates about traffic in this ODD. These predicates primary describe abstract behaviour of dynamic objects, e.g., ego vehicle follows another vehicle, as well as relations between entities of the infrastructure, e.g. the road has a speed limit, and between the infrastructure and dynamic objects, e.g., a vehicle stops at a red light.

For the evaluation of the representativity, data from real or artificial traffic scenarios is segmented based on the infrastructure, analyzed to interpret and evaluate the abstract predicates. The evaluation of predicates is gathered and compared to the possibilities of the ODD to determine the representativity of the analyzed scenario. Our methodology does not only allow for analyzing the representativity of single scenarios but also the representativity of scenario sets as well as the comparison of multiple scenarios with each other.

Funding

  • Industrial Collaboration