A Cross-Layer Optimization Technique Based on Bayesian Reinforcement Learning Extended POMDP Model for Dynamic Spectrum Access
Rongkui Zhang · 2022
Dynamic wireless environment poses a unique challenge to network communication. In the cognitive radio (CR) based broadband network, nodes not only need to detect the free frequency band through broadband spectrum sensing, but also pay attention to the necessary dynamic spectrum access (DSA) to effectively improve spectrum utilization. To solve the above problems, this paper proposes a cross layer optimization DSA framework based on Bayesian Reinforcement Learning extended POMDP model (BRL-POMDP). First, by combining MAC layer and PHY parameters, the spectrum sensing and access problems of secondary users (SU) under the condition of hardware resource constraints are modeled as BRL-POMDP; Then it is transformed into belief state Markov decision making (B-MDP) process; Finally, an algorithm based on posterior model parameter sampling is used to track the confidence level, so as to generate the optimal spectrum sensing and access strategy. The simulation results show that the algorithm can effectively improve the transmission capacity and spectrum utilization, thereby reducing the bit error rate under the conditions of power limitation and dynamic environment.