Rigid-Body Simulation-Based Probabilistic Analysis of Drone Crash Impact Points and Its Embedded System Application

Abstract

This paper presents an offline-to-embedded workflow for estimating the displacement of an unmanned aerial vehicle after total thrust loss. A C#/.NET simulator integrates rigid-body motion through BepuPhysics2 and uses Monte Carlo trials with randomized steady wind and stochastic turbulence. A grid sweep over initial altitude, horizontal speed, and vertical speed produces crash-location samples that are exported to CSV and summarized by a C99 lookup-table library. The originally reported experiment contains 62,912 trials over 64 altitude–horizontal-speed combinations and reports an overall mean displacement of 20.33 m and a sample 95th percentile of 49.02 m. The embedded library uses static buffers, grouped sample statistics, and trilinear interpolation to return estimated offsets and a radius proxy without running the physics simulator on the flight controller. Inspection of the published implementation reveals important distinctions between an empirical percentile and the normal-approximation radius used by the C library, as well as assumptions concerning interpolation, out-of-grid inputs, and simulation fidelity. The findings therefore demonstrate a computationally lightweight, simulation-based prediction pipeline, while the reported radii should not be interpreted as field-validated safety guarantees.