Effective cooperative spectrum sensing using deep recurrent reinforced learning‐based Q‐routing in multihop cognitive radio networks

Abstract

Summary Cognitive radio network (CRN) is a promising technology that mitigates the scarcity of spectrum. The main challenge faced in the opportunistic CRN is the spectrum sensing (SS). The traditional methods of SS use detectors to find the available bands, which appear to provide unreliable performance in the case of actual environment where noise is predominant. Literature has proved that employing an artificial intelligence (AI) model for routing overcomes the lack of network knowledge and lack of human intervention and helps to learn robust patterns. Reinforced learning (RL) is a dynamically learning process that selects the actions based on continuous feedback received from the dynamic environment to maximize the reward. Deep reinforcement learning (DRL) models have proven to be successful at learning control policies image inputs. However, they struggle with learning policies that require longer term information. Recurrent neural network (RNN) architectures have been used in tasks dealing with longer term dependencies between data points. Motivated by the performance of AI models, these architectures are investigated in this work to overcome the difficulties arising from learning policies with long‐term dependencies. Thus, a deep recurrent reinforced learning‐based Q‐routing (DRRL‐based Q‐routing) is developed. The proposed study suggests a multihop CRN operated in an interweave mode. This algorithm finds optimal routing path between the secondary user transmitter (SUT) and the secondary user destination (SUD), optimal SS duration, and individual secondary user (SU) power requirements for SS and data transmission process while minimizing the end‐to‐end outage under the constraints of energy causality, SS reliability, interference threshold, and individual link throughput.