Reinforcement learning enhanced iterative power allocation in stochastic cognitive wireless mesh networks

As the scarce spectrum resource is becoming over-crowded, cognitive wireless mesh networks have great flexibility to improve the spectrum utilization by opportunistically accessing the licensed frequency bands. One of the critical challenges for realizing such network is how to adaptively allocate transmit powers and frequency resources among secondary users (SUs) of the licensed frequency bands while maintaining the quality-of-service (QoS) requirement of the primary users (PUs). In this paper, we consider the power control problem in the context of cognitive wireless mesh networks formed by a number of clusters under the total transmit power constraint by each SU as well as the mean-squared error (MSE) constraint by PUs. The problem is modeled as a non-cooperative game. A distributed iterative power allocation algorithm is designed to reach the Nash equilibrium (NE) between the coexisting interfered links. It offers an opportunity for SUs to negotiate the best use of power and frequency with each other. Furthermore, how to adaptively negotiate the transmission power level and spectrum usage among the SUs according to the changing networking environment is discussed. We present an intelligent policy based on reinforcement learning to acquire the stochastic behavior of PUs. Based on the learning approach, the SUs can adapt to the dynamics of the interference environment state and reach new NEs quickly through partially cooperative information sharing via a common control channel. Theoretical analysis and numerical results both show effectiveness of the intelligent policy.