Deep Learning Based Antenna Muting and Beamforming Optimization in Distributed Massive MIMO Systems

Inspired by the success of Deep Learning (DL) in solving complex control problems, a new DL-based approximation framework to solve the problems of antenna muting and beamforming optimization in distributed massive MIMO was proposed. The main purpose is to obtain a non-linear mapping from the raw observations of networks to the optimal antenna muting and beamforming pattern, using Deep Neural Network (DNN). Firstly, the antenna muting and beamforming optimization problem is modeled as a non-combination optimization problem, which is NP-hard. Then a DNN based framework is proposed to obtain the optimal solution to this complex optimization problem with low-complexity. Finally, the performance of the DNN-based framework is evaluated in detail. Simulation results show that the proposed DNN framework can achieve a fairly accurate approximation. Moreover, compared with the traditional algorithm, DNN can be reduced the computation time by several orders of magnitude.