MORE | Fall 2025

Experimental Evaluation of Model-free Framework for Fault Detection and Identification

Fault detection is essential for maintaining the safety of unmanned aerial vehicles (UAVs). Traditional approaches rely on model-based residuals that often fail under system uncertainties. This research evaluates a model-free neural framework for real-time fault detection and identification (FDI) in quadrotor systems. The framework eliminates dependence on mathematical models by learning fault patterns directly from data. Simulation and hardware experiments using Crazyflie drones with induced motor faults will evaluate detection accuracy, latency, and recovery performance, advancing safer and more resilient autonomous flight control.