On the Performance of Multihop D2D Communications in Cognitive IoT Networks with Energy Harvesting: A Deep Learning Approach

In this paper, we study multi-hop device-to-device (D2D) communications in cognitive Internet-of-Things (IoT) networks, where all D2D devices harvest energy from a power beacon through energy beamforming to transmit their signals to a destination with the existence of a multiple antennas primary receiver. Under Nakagami-m fading channels, the exact and asymptotic expressions for the outage probability (OP) of the considered system are derived then validated by Monte Carlo simulations. We then develop a deep neural network (DNN) for the OP evaluation with a small root-mean-square error. The designed DNN model can predict the OP with high accuracy while it drastically reduces the execution time, expediting a real-time configuration for multi-hop D2D communications in cognitive IoT networks with energy harvesting.