This paper proposes a concept for the semantic classification of pedestrian traffic scenarios to estimate the test coverage during the validation of automated driving systems for urban environments. A model is introduced to describe pedestrian behavior as a combination of geometric data of their trajectory and maneuvers in relation to the surrounding infrastructure. The presented concept is applied and evaluated on an initial dataset of recorded measurement data to monitor the achieved coverage of pedestrian behavior during the test drives. Furthermore, the benefits and challenges of the approach are discussed. The methodology can help to detect semantic redundancy in a given scenario database and thereby improve the efficiency of the overall testing process. Moreover, common traffic scenarios can be identified, which makes it possible to focus real world test drives on these, while more infrequent scenarios are covered in simulations.
Semantic Classification of Pedestrian Traffic Scenarios for the Validation of Automated Driving
2019-10-01
232211 byte
Conference paper
Electronic Resource
English
Semantic classification of urban traffic scenarios for the validation of automated driving systems
TIBKAT | 2023
|