Deep Learning-Based Spectral and Energy Efficiency Optimization for CoMP in HetNets.

In this paper, we consider a multiple-input single-output heterogeneous network (HetNet) with a macro base station (BS) and multiple small BSs. Assuming that the BSs employ coordinated multi-point (CoMP) transmission, we propose the deep neural network (DNN)-based dynamic cell selection and power allocation scheme for the CoMP transmission. In the proposed scheme, the DNNs are trained with predefined loss functions to maximize spectral efficiency and energy efficiency. Through simulation, we show that the proposed DNN-based scheme can achieve similar performance to the optimal algorithm but with significantly lower complexity. In addition, we introduce a CoMP threshold, which corresponds to the number of CoMP BSs, to adaptively adjust the CoMP complexity of the proposed DNN-based scheme. It is observed that the trade-off between spectral and energy efficiencies and complexity can be exploited by pertinently choosing the CoMP threshold.