FD-NOMA-enabled UAV-vehicle collaborative networks: Channel capacity analysis with imperfect CSI under Rayleigh fading

The continuous growth in the number of vehicles has led to increasingly scarce spectrum resources. However, uncrewed aerial vehicles (UAVs), with their flexibility and mobility, combined with full-duplex non-orthogonal multiple access (FD-NOMA) technology, form a UAV-vehicle collaborative networks that offers a potential solution for improving spectrum efficiency. Influenced by the mobility of UAVs and vehicles, it is crucial to study how to quickly and accurately analyze the total channel capacity. Therefore, we derive closed expressions and approximate solutions for the total channel capacity in FD-NOMA-enhanced UAV-vehicle collaborative networks. In addition, considering the difficulty of accurately obtaining channel state information (CSI) in real time, a deep learning-based CSI estimation method is designed. By incorporating least square (LS) coarse estimation, deep neural network (DNN) denoising, bidirectional long short-time memory (BiLSTM) time-domain prediction, and weighted dimensionality reduction processing, the estimation accuracy in high-speed scenarios is significantly improved. Finally, the simulation results show that the capacity of the constructed FD-NOMA system in the low signal to noise ratio (SNR) region is improved by about 1.8–2.5 bps/Hz compared with that of full-duplex orthogonal multiple access (FD-OMA), and the CSI estimation error based on deep learning is reduced by 85% compared with that of the traditional LS algorithm. In addition, stable channel capacity is maintained at vehicle speeds up to 80 km/h.