The evolution of 5G networks over the last few years has introduced a variety of technologies for more efficient radio access networks (RANs), which end up in ultra-dense heterogeneous infrastructure with deployments of high complexity. In this paper, we propose a new framework for RAN design in ultra-dense urban scenario based on the machine learning. The key idea of the proposed framework is to bring intelligent capabilities to the coverage planning problem for complex multi-tier scenarios, in order to achieve better network performance. We design our framework for small cells coverage optimization with 3D urban environment, macro cell locations, and realistic traffic statistics. Simulation results show that our proposed intelligent RAN framework significantly outperforms the conventional coverage design solutions, even after only a short learning time.
Read full abstract