automotive-e2e-safety-analysis

What problem does it solve? End-to-end autonomous driving systems that use neural networks for the full perception-to-control pipeline cannot be verified with traditional modular safety methods, making ISO 26262 compliance, SOTIF analysis, and failure mode attribution extremely difficult. ## Core Features & Use Cases - Safety Architecture Patterns: Provides safety cage and dual-path (1oo2D) architecture designs that wrap E2E neural networks with ASIL-rated rule-based monitors and fallback behaviors. - SOTIF Analysis for DNNs: Identifies DNN-specific triggering conditions such as out-of-distribution inputs, adversarial perturbations, distribution shift, and model uncertainty, with corresponding validation approaches. - ISO 26262 Compliance Strategy: Defines ASIL decomposition at the system level, safety mechanism specifications (collision monitor, kinematics check, ODD monitor, model health), and safe state definitions. - Use Case: A safety engineer designing an E2E ADS can use this Skill to produce a safety architecture review, an ASIL decomposition strategy, and a scenario-based V&V plan exceeding 10M km of equivalent simulation. ## Quick Start Ask the agent to design a safety cage architecture with ASIL decomposition and SOTIF triggering condition analysis for an end-to-end autonomous driving system.