A Data-Driven and Load-Aware Interference Management Approach for Ultra-Dense Networks

Abstract

The ultra-dense network (UDN) has been widely accepted as a promising technology to improve the network performance. However, the severe co-channel interference (CCI) generated due to the densely deployed femtocells greatly limits the network throughput. Different from most conventional methods that model the inter-user interference intensities based on the accurate geographical distance information, which is usually hard to obtain in reality, a more practical machine learning based relative interference intensity modeling method is proposed. The proposed method models the relative interference intensities by mining the resource block (RB) allocation data, the new data indicator (NDI) data, the acknowledgement (ACK) and negative acknowledgement (NACK) data collected from the network, which could achieve an extremely high accuracy that is validated by the simulation results. In addition, we propose a load-aware resource allocation approach which calculates each user's boundary of reusing the common RBs and allocating the orthogonal RBs with its interfering sources based on the relative interference intensities modeled above and the network load in each transmission time interval (TTI). The orthogonal interfering source set of each user is generated based on its time-varying boundary. Simulation results show that the proposed load-aware resource allocation approach outperforms all the benchmark algorithms under most network densities and network loads especially when the network load is heavy and the network is ultra-dense.

Keywords

• Ultra-dense network
• machine learning
• association rules
• relative interference intensities
• acknowledgement and negative acknowledgement