Name of the participant: Igor Vozniak
Description of the IT research project: Autonomous driving domain and related challenges gained a lot of attention in recent years. Despite impressive achievements, the goal of reaching a fully automated driving (L5) as in accordance to SAE classification remains a challenge, where the key components towards the successful deployment of those systems remain the same, namely safety of the passengers of autonomous vehicles as well as the other participants of the traffic, e.g. pedestrians, other vehicles. This project aims to make another step towards certified autonomous vehicles (L4-L5), where we follow a common system validation technique, by shifting the real problem to a synthetic domain, e.g. simulated environments like OpenDS.
Moreover, we propose another view on critical scenario generation challenge from the perspective of static elements of the simulated scene. According to our scientific hypothesis, the geometric/architectural complexity of the applied buildings must be considered and to be part of the scene complexity metrics for further validation of adaptive driving assistance systems (ADAS) of the vehicle. Throughout the workflow a novel pipeline will be implemented, where we propose openSCENE as an addition to existing open formats, namely openDrive, openCRG, openSCENARIO, OSI, which are broadly applicable in the process of critical scenario generation.
The main contribution of openSCENE project, is the proposed toolchain, capable of automatically reconstructing highly realistic copies of buildings in 3D formats (mesh objects), using AI methods and unstructured point cloud data (PCD) as the only input. Moreover, we introduce an extension, namely the generation of “unseen” 3D models of buildings with a manually set level of architectural similarity. The ultimate goal of the project is to complete the concept of digital reality cycle (by shifting the challenge to simulated environment) for critical scenario generation which is essential for evaluation and certification of autonomous vehicles by invoking the introduced in the scope of this project criticality metrics. The purpose of which is to detect the relation between the used static buildings of certain architectural style and the criticality of the traffic scenario. OpenSCENE project is supported by FARO company (data partner), the innovative industry leader for high-definition LIDAR sensors and point cloud post-processing, by providing the ground truth data.
Software Campus partners: DFKI, Huawei
Implementation period: 01.01.2021 – 31.12.2022
