Robust and comprehensive resource distribution for random access networks with imprecise CSI

Abstract

For random access based wireless networks, where the interference channel state information (I-CSI) and the communication channel state information (C-CSI) were both unpredictable, a comprehensive and strong resource allocation paradigm was presented.The suggested resource allocation framework used deep neural networks (DNNs) to approximate the ideal resource allocation policy in an unsupervised way, treating the optimization goal of wireless networks as a learning issue.Two concatenated DNN units—the power control unit and the uncertain CSI processing unit—were developed by representing the CSI uncertainties as ellipsoid sets.The two concatenated DNN units were then jointly trained using an alternating iterative training approach.In conclusion, the simulations confirm that the suggested robust leaning technique is more effective than the nonrobust one.