Reproducibility Package for the Risk-Aware AI Architecture for BVLOS UAV Safety (RASA)

This reproducibility package contains the mathematical parameters, risk-weighting coefficients, visualisation assets, and supporting metadata for the Risk-Aware AI Architecture for BVLOS UAV Safety (RASA). The repository documents the core risk quantification model: R(t) = α·U_sensor(t) + β·L_c,norm(t) + γ·U_sensor(t)·L_c,norm(t) where U_sensor(t) represents normalised sensor uncertainty, L_c,norm(t) represents normalised SATCOM communication latency, and γ represents the nonlinear interaction coefficient capturing compound failure effects. Included materials comprise: • Figure 1 – RASA system architecture and operational workflow • Figure 2 – Risk surface visualisation illustrating sensor uncertainty and communication latency interactions • Figure 3 – Monte Carlo simulation results for representative BVLOS operational scenarios • Graphical Abstract (GA) summarising the complete three-layer RASA framework • Metadata documentation describing model assumptions, parameter definitions, and risk-state classifications The package supports transparency, reproducibility, and independent verification of the RASA framework for Beyond Visual Line of Sight (BVLOS) UAV operations under latency-constrained satellite communication environments.