A Neural Network Prediction-Based Adaptive Mode Selection Scheme in Full-Duplex Cognitive Networks

In this paper, we propose a neural network (NN) predictor for multi-slot prediction and an adaptive mode selection scheme, with the goal of improving secondary users (SUs) throughput while alleviating collision to primary user (PU) in full-duplex (FD) cognitive networks. Conventionally, FD SU can either operate in a transmission-and-reception (TR) mode to improve its throughput, or a transmission-and-sensing (TS) mode to avoid collision to PU. The difference between TR and TS modes in goal gives rise to a trade-off between higher SUs throughput and lower collision probability, which can be optimized by allowing SU to switch between these two modes. Accordingly, we design an NN predictor to predict PUs future activity which is considered as the basis of switching. In such a context, we analyze the prediction performance in terms of prediction error probability. We also compare the performance of our proposed scheme with conventional TR and TS modes in terms of SUs average throughput and collision probability, respectively. Simulation results show that our proposed scheme achieves almost the same SUs average throughput as TR mode when PU has low tolerance for collision. Meanwhile, the collision probability can be reduced by up to 92% close to that of TS mode.