AV_Safety

Quantifying collision risk to answer: "How safe is safe enough for autonomous vehicles?"

An independent research portfolio on autonomous vehicle safety, built with evidence-first rigor. Grounded in real-world crash data, Bayesian analysis, and international safety standards.

Research Focus

• Quantitative safety thresholds for AV deployment
• Collision risk modeling using statistical and Bayesian methods
• Standards alignment with UL 4600, ISO 21448 (SOTIF), ISO 26262, NHTSA guidance
• Cross-jurisdiction analysis (USA, Canada, UK)

Project Structure

AV_Safety/
├── SKILL.md          — Master skill (project overview & skill tree)
├── docs/             — Research, standards, architecture
├── data/             — Raw and processed datasets
├── src/              — Analysis, risk models, standards, data pipeline
├── ui/               — Portfolio visualizations and dashboards
├── models/           — Trained models and checkpoints
├── tests/            — Unit and integration tests
├── notebooks/        — Exploratory analysis
├── scripts/          — Reproducible analysis scripts
├── config/           — Configuration
├── docker-compose.yml
└── requirements.txt

Working Principles

1. Evidence-first — No assumptions. Every claim backed by data or verified sources.
2. Reproducible — Every analysis produces verifiable, repeatable results.
3. Rigorous — Statistical methods documented and justified.
4. Traceable — All data sources, transformations, and decisions versioned and logged.

Skill Tree

Capability	Status
Project Setup	✅ Done
Standards Research	🔲 Pending
Data Ingestion	🔲 Pending
Risk Metrics	🔲 Pending
Bayesian Analysis	🔲 Pending
Collision Modeling	🔲 Pending
Safety Thresholds	🔲 Pending
Visualization	🔲 Pending
Portfolio Deploy	🔲 Pending

Quick Start

# Setup environment
cd AV_Safety
pip3 install -r requirements.txt

# Run exploratory analysis
jupyter notebook notebooks/

# Build dev environment
docker-compose up -d

Standards & Regulatory Framework

• UL 4600 — Standard for UAS Destination Guidance and AV safety
• ISO 21448 (SOTIF) — Safety of the Intended Functionality
• ISO 26262 — Road vehicles functional safety
• NHTSA AV Guidance — Federal autonomous vehicle safety guidance
• UK AV regulations — Transport England and DMV equivalent frameworks

License

© 2026 Ahmed. Independent research project.

---

Built with rigor, not assumptions.