Online Energy Efficiency Learning for Uplink Transmission in Dynamic H-CRAN

Dynamic changes of the system cannot be neglected in the design of the energy-efficient wireless systems, especially, in heterogeneous cloud-radio access networks (H-CRAN), where inter- and intra - tier interference are co-existing. Confronted with this problem, we investigate the online learning of energy efficiency (EE) optimization problem in dynamic systems during the uplink transmission. Specially, we focus on the dynamic H-CRAN system, where channel characteristics change arbitrarily over time slots. Therefore, we formulate the online EE problem including joint optimization of the macro base station (MBS) users (MUEs) and small cell radio remote head (RRH) users (RUEs). To tackle the formulated EE optimization problem, we propose a simple and projection-free online optimization policy, namely online Frank-Wolfe (OFW) method, which allows users to outperform the best dynamic transmit policy, despite changes of channel characteristics over the given time slots. The proposed projection-free OFW method is capable of obtaining optimal solution for EE maximization problem and achieve Nash equilibrium of the underlying game. Moreover, the regret bound expression for the OFW method is derived. Numerical results confirm the theoretical analysis and demonstrates the superiority of the proposed OFW method compared to the benchmark method.