Deep-Learning-Based System-Level Energy Efficiency Maximization in Ultra-Dense Micro Cell Networks

The energy-efficiency enhancement is crucial for the resource management in the practical deployment of the ultra-dense micro cell network (UDMCN). This letter proposes a deep learning based resource management framework for the integrated designment of transmission power and user association in the UDMCN. Specifically, the proposed framework essentially belonging to unsupervised learning architecture is composed of three distinct deep neural network (DNN) units, each of which is responsible for transmission power controlling, user association and switching ON/OFF decision of the micro base station (MBS) respectively. Simulation demonstrates that the proposed framework can achieve near-optimal performance with acceptable computing time.