A Multi-Modal AI System for Detecting Pedestrians Lying on the Road: Simulation-Based Safety and Injury Risk Analysis

⚠️ This record has been superseded. Please use the current version: https://doi.org/10.5281/zenodo.20041467 # **v1.1.0: Technical Supplementary Materials — Simulation & Risk Data** This repository contains the technical implementation and supplementary data supporting the 2026 flagship research paper: **"A Multi-Modal AI System for Detecting Pedestrians Lying on the Road: Simulation-Based Safety and Injury Risk Analysis."** #### **🔧 Key Technical Components**1. **Forensic Impact Model Scripts**Implementation of the analytical model linking detection latency ($t_d$) to injury outcome:- **Stage 1 (Kinematics):** $v_{impact} = \max(0, v_0 - a \cdot (t_{avail} - t_d))$- **Stage 2 (Biomechanics):** $HIC = k \cdot v_{impact}^{2.5}$- **Stage 3 (Clinical Risk):** $P(AIS \geq 5) = \frac{1}{1 + \exp(-\alpha + \beta \cdot \ln HIC)}$ 2. **ISO 26262 HARA Data**Structured Hazard Analysis and Risk Assessment (HARA) datasets supporting **ASIL D** classification for fallen pedestrian detection at night. 3. **Supplementary Figures (v1.1.0 New)**- **Figure S1:** Monte Carlo Uncertainty Propagation ($N = 100,000$).- **Figure S2:** Tornado Plot (One-Way Sensitivity Analysis).- **Figure S3:** Continuous Injury Risk Curve (HIC vs. $P(AIS \geq 5)$). #### **📊 Results Summary**The AFODS framework achieves a **98.2% True Positive Rate (TPR)** under clear daytime conditions. Monte Carlo analysis confirms that this intervention reduces the probability of fatal head injury from a **76.5% baseline** to **0.4% (95% CI: 0.0–3.7%)** in worst-case night/rain scenarios. Related Publication (Architecture)Barua, N. (2026). Integrated Safety Architectures: Leveraging Multi-Modal AI and ISO 26262 to Protect Vulnerable Road Users. SSRN.https://dx.doi.org/10.2139/ssrn.6112086 Related Publication (Baseline)Barua, N.; Hitosugi, M. (2025). Advanced Multi-Modal Sensor Fusion System for Detecting Falling Humans. Vehicles.https://doi.org/10.3390/vehicles7040149 Intellectual Property DisclosureSystem architecture protected under Japanese Patent Application No. 2025-167440 (filed 3 October 2025).