Performance of Deep Learning Methods for Different Relay Selection Protocols

Abstract

Deep Learning (DL) is an emerging topic that has found its place in various applications due to its many advantages, such as providing good performance and reducing complexity. In this paper, we employed different DL methods for channel estimation in cooperative communication by using Maximum Likelihood Estimation (MLE) to combine different copies of transmitted signal over a Binary Phase-Shift Keying (BPSK) modulation in a Rayleigh fading communication channel. We have presented three different relay selection protocols. The first one is SNR-based; the SNRs of the transmission lines between nodes are calculated, and a relay with maximum SNR is selected among the relays. The second relay selection protocol is threshold-based; average SNRs have been computed for all transmission lines, and the links exceeding the average SNR attend in the collaboration. The last protocol is that all relays in the network participate in cooperation, which achieves better performance metrics in terms of Bit Error Rate (BER) and outage analysis. From our simulation results, DL algorithms achieved good performance for channel estimation in the different relay selection protocols based on the Decode and Forward (DF) relaying system.