Mid-term status report on KISSaF: AI-based situation interpretation for automated driving

KISSaF is a publicly funded project with four project partners from industry and academia. The aim of project KISSaF is the development of a robust scene prediction model for automated driving. State-of-the-art Deep Learning methods are used for a complete and reliable forecasting of the traffic scene with large time horizons. The underlying environment modeling uses a graph-based representation of the scene. A prototype vehicle has been built-up for data recording. This data is the central part for model development, improvement and testing. A framework is currently setup for a scenariobased test approach and performance can be judged under realistic conditions with integrated maneuver planning.