Intercell-Interference Cancellation and Neural Network Transmit Power Optimization for MIMO Channels

Recent demand for capacity increase in mobile communications urges the development of both multiple- input multiple-output (MIMO) and small cells. However, overlapping coverage areas between neighboring small cells results in interference-limited networks, where the capacity of the network is severely degraded without superior intercell interference management (IIM) techniques. As one of the superior IIM techniques, the power control intercell interference coordination (ICIC) optimizes downlink transmit powers in the network. This paper proposes an IIM scheme consisting of both neural networks (NNs) power control on the transmitter side and interference cancellation (IC) on the receiver side. Computer simulations examine networks of MIMO systems with the power optimization using some ICIC algorithms: NN, the greedy search, the belief propagation (BP), and the maximum power, combined with IC on the receiver side. In addition, the deep learning, a breakthrough technique for NN with multi-layer structures, is also applied to mobile communications. The results show that the performance of NN is very close to that of the greedy search and superior to that of BP. Complexity of NN is much less than that of BP, and thus NN is suitable for IIM. It is also demonstrated that employing IC provides high capacity gain.