Deep Neural Network Based Cell Sleeping Control and Beamforming Optimization in Cloud-RAN

Abstract

Cloud Radio Access Network (Cloud-RAN) is a promising network architecture for the next generation (5G) wireless communication. Despite the remarkable benefits on the capacity, the provision of power-efficient wireless resource management solution is still challenging. Cell sleeping control and cooperative beamforming design are considered as two enabling ways to address this issue. In this paper, we propose a novel Deep Neural Network (DNN) based approach to minimize the power consumption of network while satisfying the QoS demand by jointly designing the RRHs sleeping modes and beamforming weights inside a certain cell. The key idea of proposed DNN-based approach is to learn the non-linear mapping from channel coefficients to the optimal beamforming weights and the sleeping modes of multiple RRHs. Compared with the conventional numerical optimization schemes which require extensive iterations and result in considerable computational complexity and limited application in real-time processing, the proposed DNN-based approach enables the real-time cell sleeping control and beamforming optimization. Simulation results show that the DNN- based approach improve the energy efficiency significantly and calculative efficiency about three orders of magnitude.