Vision-based Gyroscope Fault Detection for UAVs

Abstract

This paper presents a vision-based fault detection and isolation architecture for unmanned aerial vehicles. The vehicle’s attitude is computed from visual input through a horizon tracking algorithm, independently of any other sensor. In a second stage, two Kalman filters are used for fault detection and identification in two gyroscopes. The loosely coupled architecture is suitable for real-time application. The algorithm was implemented with the ROS framework and the system’s performance is evaluated in a real-time application scenario with artificially introduced sensor faults.