Autonomous Driving Safety, Simulation & Validation Framework: Interface-Level Failure Propagation Analysis with SOTIF Scenario-Based Testing and Closed-Loop Verification
Year: 2026
Type: article
Abstract: A simulation-based safety validation program for autonomous driving systems following the automotive V-model — Specify → Integrate → Execute → Evaluate. Covers ISO 26262 Hazard Analysis and Risk Assessment (HARA), SOTIF (ISO 21448) scenario classification (T1-T5), closed-loop simulation in CARLA 0.9.15, KPI measurement, coverage analysis, and a GSN safety case with full evidence traceability. Key results: 160 closed-loop runs across 8 SOTIF scenarios, 3/5 safety goals verified, zero collisions with uncertainty-aware planner active. Key finding: sensor trust reweighting (Loop 1) provides zero standalone safety benefit — architecturally dependent on uncertainty-aware planning (Loop 2). Extended with interface-level failure propagation analysis measuring how sensor failures traverse the perception-planning pipeline. Failure Propagation Coefficient (FPC) metric developed and validated across 45 injection runs. Methodology validated against Foretellix Safety-Driven V&V Guide (2024).
Author: Vaghela, Nihar
Institution: Friedrich-Alexander-Universität Erlangen-Nürnberg
Language: English
Subfield: Safety, Risk, Reliability and Quality
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7165018478
FWCI:
Cites:
Cited by:
Related to: