Using Machine Learning Algorithms as part of Reliability Analysis Workflows for Digital Validation of Advanced Driver Assistance Systems
Digital event-based validation is a key element to validate and verify L3 and higher level of automated driving systems. The presentation addresses challenges of digital validation using reliability analysis approaches, like fitting appropriate distribution functions for uncertainty definition, understanding which uncertainties will dominate the risk, using machine learning for sensitivity analysis and metamodeling and using advanced reliability analysis methods for reliable estimates of risks. Workflows of reliability algorithms and machine learning will be introduced which are used by our customer as one pillar to successful certify ADAS Level 3 systems.
- Digital Verification of automated driving systems
- Uncertainty Definition is a key for quantification of safety
- Importance to understand the driver for response variability
- How to overcome the exorbitant costs of Monte Carlo Sampling quantifying rare event probabilities