Back to Program

Training and Testing Autonomous Vehicles on Edge Case Scenarios

Dr. Nils Goldbeck

CTO

∂RISK

We show how AV developers can use a knowledge graph technology to access vast amounts of edge case scenarios based on real-world data and failure mode analysis. ASAM OpenSCENARIO and OpenDRIVE provide a standardised interface between the knowledge graph and various simulators, including NVIDIA DRIVE Sim, rFpro and CARLA. In case studies, we demonstrate how edge case scenarios have been used for requirements-based testing of sensors, training object detection models, improving motion planning algorithms and ODD risk assessment.