Channel Estimation and Relay Selection for Space–Air–Ground–Sea Integrated Networks
Reliable communication over oceans and in remote areas remains challenging because most wireless networks depend on land-based infrastructure. Space–air–ground–sea integrated networks (SAGSINs) provide a promising solution by integrating space, aerial, terrestrial, and maritime network components to extend coverage beyond the reach of land-based networks. In this paper, we consider a maritime relay-assisted SAGSIN where a sea-surface station (SS) communicates with a base station (BS) through one relay selected from three candidate platforms: an onshore station, a high-altitude platform, and a satellite (SAT). Since these relay links operate in different propagation environments and network segments, accurate channel estimation and relay selection become challenging. The proposed framework considers least squares (LS), linear minimum mean square error (LMMSE), and an adapted denoising convolutional neural network (DnCNN)-based estimator for channel estimation over heterogeneous maritime relay links. The DnCNN-based estimator learns the nonlinear mapping between the initial channel estimate and the corresponding refined channel estimate, thereby reducing estimation errors caused by noise and limited pilot observations. The refined channel estimates are then used to support relay selection, so that the SS can choose a suitable relay for forwarding its data to the BS. The simulation results confirm that the adapted DnCNN-based estimator generally provides lower normalized mean square error than the traditional LS and LMMSE estimators, especially at low transmit power and short pilot length. The results further show that the proposed relay selection method achieves an end-to-end data rate close to the perfect channel state information benchmark. These results confirm that accurate channel estimation improves relay selection and enhances maritime communication performance.