“…But the conventional algorithms cannot satisfy the increasing coverage because UAV movements require much energy and slow navigation deteriorates the coverage performance. Recently, by exploiting the potentials of machine learning into wireless communication, a deep learning-based wireless communication method provides an alternative mean for optimizing the UAV navigation problem, whose performance has been corroborated in non-orthogonal multiple access (NOMA) [9], massive MIMO [10], [11], traffic control [12], [13], routing techniques [14], software defined network (SDN) [15], UAV [16], [17], and millimeter-wave (mmWave) communication [18], etc.. In particular, [17] proposed a deep learning-based method for UAV navigation without requiring sensing data that provides mapping information, however this method cannot converge quickly, which makes it difficult to be applied in real-time navigation scenarios.…”