An Adaptable UAV Sensor Data Anomaly Detection Method Based on TCN Model Transferring

Unmanned aerial vehicles play a critical role in both military and civilian applications, and their safety and reliability have also been paid more and more attention. UAV anomaly detection can detect and eliminate potential faults in a timely and effective manner, reducing the probability of accidents. Due to the influence of the complex and changeable operating environment, data shifting problems are inevitable in time series anomaly detection. Ignoring this issue may result in a significant drop in the accuracy of anomaly detection. Therefore, a UAV sensor data anomaly detection method based on Temporal Convolution Network (TCN) model transferring is proposed in this paper. First, the TCN model is pre-trained by using a large amount of data in the source domain. Then, parameters of the model are fine-tuned on the target domain. Finally, the threshold detection method is used to determine whether there is abnormality in the UAV sensor data. This work aims to address the multiple modes of UAV and improve the data-driven adaptivity for anomaly detection. In the experiments, the flight sensor data are used to verify the performance of the proposed model. The results show that the proposed method achieves high precision, high detection rate and low false detection rate in different domains.