Intelligent Resource Management for Satellite and Terrestrial Spectrum Shared Networking toward B5G

Integrated satellite-terrestrial networks (ISTNs) toward beyond fifth-generation (B5G) wireless systems benefiting from both satellite and terrestrial systems can achieve all-time seamless and broad coverage. Considering the scarcity of frequency resources and intense satellite-terrestrial cochannel interference, intelligent resource allocation with high spectrum efficiency and low co-channel interference has received a substantial amount of attention. Focusing on the spectrum efficiency advantages achieved by spectrum sensing and prediction, a hierarchical satellite and terrestrial spectrum shared framework based on the spectrum management unit (SMU) is proposed. Moreover, an intelligent resource management scheme in the SMU composed of spectrum sensing, prediction and allocation is formulated to improve spectrum efficiency with different user densities. We present a support vector machine (SVM) based algorithm that improves the accuracy and robustness of the learned model for the detection of spectrum occupancy. Then, a convolutional neural network (CNN) based spectrum prediction (SP) is performed, where the CNN is trained with the historical detection results from spectrum sensing. In addition, an intelligent resource management scheme including spectrum sensing, prediction and allocation based on the priorities and requirements of users is proposed to improve spectrum utilization. The evaluation results demonstrate that the proposed intelligent resource management scheme can achieve lower error detection probability and better spectrum efficiency.