Deep Learning Based Throughput Estimation for UAV-Assisted Network

Due to the rapid growth of mobile technology, unmanned aerial vehicles(UAV) is emerging as a promising solution to distribute wireless data for ground users as a base station (BS). Our study focuses on the analysis of UAV-based BS that assist aerial wireless network. We practically used the real data measurement from the UAVs connected with ground mobile users as air-to-ground(A2G) communication service. The main aim of our work is to analyze and estimate the UAV-BS user throughput with different parameters such as height and distance. In order to achieve our objective, we have estimated the locations of UAVs' and mobile-users', heights of UAV, elevation angle and nature of LoS/NLoS. The system performances are evaluated through long short term memory(LSTM) and comparison was made with multi-layer perceptron(MLP) algorithm. Finally, the evaluation result shows the system has accurate and motivated prediction performances of the user throughput.