On-line and non-invasive anomaly detection system for unmanned aerial vehicle

Abstract

There are almost no on-board intelligent anomaly detection systems in most of the existing unmanned Aerial Vehicles (UAVs), and the flight status assessment still depends on ground control station. While, this method can't meet the requirement of real-time anomaly detection for UAV autonomous and safe flight. In order to achieve real-time monitoring of UAV flight status, and improve the reliability and safety of UAVs. In this paper, an on-line and non-invasive embedded anomaly detection system (EADS) is designed to solve the problem of UAV on-board anomaly detection in complex environment. During the flight, whether the sensors and hardware components are abnormal or not is continuously detected via flight data. The proposed embedded anomaly detection system is divided into two parts, (1) hardware platform for heterogeneous calculation is based on Xilinx Zynq-7000 SoC with dual-core Cortex A9 processors and Field Programmable Gate Arrays (FPGA); (2) on-line anomaly detection is based on least squares support vector machine (LS-SVM) algorithm, including flight data preprocessing and analysis. The flight data from Flight Gear is utilized for the demonstration of EADS, and the experiment results show that the proposed system is effective for UAV real-time anomaly detection.