Homotopy-Informed LSTM for Efficient Spectrum Prediction in Uav Communications

As wireless communication and unmanned aerial vehicles (UAVs) continue to advance steadily, the use of UAVs has become increasingly widespread. However, the scarcity of spectrum resources available for UAV communication poses a significant challenge. Spectrum prediction technology emerges as a key solution to alleviate this problem. Traditional spectrum prediction methods struggle to accurately forecast the spectrum at the next location due to the fast flight speed of UAVs and the dynamic spectrum environment. Therefore, this paper introduces a novel approach based on homotopy theory and the Long Short-Term Memory (LSTM) net-work model concept. Through experimental validation with actual measurement data, we confirmed that utilizing homotopy theory can streamline the network construction process and reliably yield good predictive results even with reduced training time.