Detecting Anomalies in Unmanned Aerial Vehicles via the Optimization Method

This study investigated the usability of optimization for anomaly detection of unmanned aerial vehicles (UAV). This research detects anomalies via the particle swarm optimization (PSO) method, focusing on the motor and blade faults. The vibration of fault data was measured via acceleration. Combining the PSO method with the monitoring-based fitness function identified the exact place where fault had happened. The vibration velocity increased two-times from the usual velocity when the fault was detected. The fitness function was developed via three stages, i.e., frame setting, tolerance checking, and computing the PSO standard to differentiate among faulty, turning, and usual data peaks. This study achieved a high detection accuracy of 76% using simulation programs of mission planner, ardupilot, and flight gear.