Anomaly detection in unmanned aerial vehicles flight data: A survey

UAVs (unmanned aerial vehicles) have become one of the most important technologies. industries, such as military, agriculture, environmental monitoring, and delivery, are based on them. The lack of a human pilot and the strong reliance on sensor data can create problems for these devices. This leads to reduced performance, greater crash risks, and potential cybersecurity threats. The process of the analysis of data to identify outliers or unusual patterns in flight data is known as anomaly detection in UAV flight data. In this review, we compare statistical methods and AI techniques for detecting anomalies in flight data from UAVs. Statistical methods like principal components analysis (PCA), regression models, and Mahalanobis distance are used to find flight anomalies. These methods are simple and efficient to use, but they have limits. They struggle with complex and non-linear patterns. AI methods like machine learning and deep learning perform better on large and complex data. They can correctly detect many types of anomalies, like point, drift, and mixed anomalies. This paper reviews past studies. It also highlights challenges and suggests future research directions in UAV anomaly detection. This study shows that using a mix of methods, hybrid learning, and better algorithms can boost the accuracy and reliability of anomaly detection systems.