Deep Learning Architecture for UAV Traffic-Density Prediction

The research community has paid great attention to the prediction of air traffic flows. Nonetheless, examining patterns for unmanned aircraft management (UTM) is relatively sparse at present. Thus, this paper proposes a one-dimensional convolutional neural network and encoder-decoder LSTM framework integrate flow with intrinsic complexity metric. This adapted metric takes into account important differences between ATM UTM operations, such as dynamic structures airspace density. Additionally, proposed methodology been evaluated verified in simulation scenario environment, which drone delivery system that considered essential COVID-19 sample tests, package services from multiple post offices, an inspection railway infrastructure fire-surveillance tasks. Moreover, model also considers impacts other significant factors, including emergency static no-fly zones (NFZs), variations weather conditions. results show achieves smallest RMSE value all scenarios compared approaches. Specifically, error 8.34% lower than shallow (on average) 19.87% regression on average.