Regret Matching Learning Based Spectrum Reuse in Small Cell Networks

We investigate the interference-aware spectrum reuse for heterogeneous small cell networks (SCNs), by specially considering dense-user deployment and stochastic-environment uncertainties. Most existing approaches, which are lack of evolving coordinations and rely on precise channel state information, tend to be inefficient in the context of dense SCNs with uncertainties. To improve the performance, by introducing a reliable metric of successful transmission probability to characterize the individual utility, we adopt a correlated equilibrium (CE)-based game to formulate spectrum reuse, and propose a distributed regret-matching learning algorithm to achieve the CE solutions. Eliminating the dependence on definite information and with general CE points consideration, our new scheme is feasible under the varying environment and can obtain more promising solutions than the state-of-art reinforcement learning methods by encouraging players to coordinate their strategies. Numerical simulations demonstrate the advantages of our proposed scheme.