Intelligent Joint Resource Management of Shared Spectrum HetNets

Shared Spectrum networks are a means of providing affordable connectivity to digitally excluded regions. Equal priority radios/AP/BS/nodes with similar and dissimilar MAC protocols in shared spectrum homogeneous and heterogeneous networks compete for scarce spectral resources. Coordination of these resources is challenging, especially with heterogeneous networks (HetNets). The task of achieving and maintaining reduced contention among requesting radios is worsened when many dynamically located radios request few available resources. To address this, we adopt an IEEE 802.19 coexistence framework and a Deep Q-Network (DQN) reinforcement learning (RL) algorithm for joint central coordination, thus improving the Quality of Service (QoS) experienced by equal-priority sharers. A central joint optimization algorithm (JOA) was built, trained, and tested. The results show that the novel model’s resource assignment averaged 98%, similar to the exclusive(fixed) resource assignment. The model’s DQN policy outperformed random, recursive, and a two-staged Q-learning allocation, as it consistently assigned resources that enabled requesting nodes’ receivers to provide a good quality of service. The results indicate significant potential for automated spectrum coordination, and efforts are ongoing to enhance the optimization performance of vast state-space DQN.