PRACTICAL IMPLEMENTATION OF THE SOFTWARE DEFINED 5G MOBILE NETWORK BASED ON CLOUD-RAN AND SDR TECHNOLOGIES

Abstract

The paper is devoted to the practical implementation aspects of a softwaredefined mobile network based on the Cloud-RAN architecture and universal software radio peripheral National Instruments USRP 2900. We propose a multilayer network architecture, which consists of a radio access network (RAN) plane, a core plane, a control plane, an artificial intelligence (AI) plane, and a monitoring system that collects data about network performance. The RAN plane provides all functions related to channel scheduling, data encoding and signal processing and combines all macro and small cells, as well as Wi-Fi access points. The core plane provides functionality of the evolved packet core such as service and packet gateways, mobility management entities and edge routers. The control plane is responsible for load balancing over all cells, spectrum sharing, spectrum reallocation, infrastructure reconfiguration, and other management parameters. The AI plane is responsible for inference and decisions making about optimal network configuration taking into account instantaneous network conditions. Such decisions are supported by a set of machine learning models, which are trained over the available statistics, collected through a long-term period of network operation. External monitoring system provides the needed data for AI plane by using the MQTT (Message Queue Telemetry Transport) protocol to collect information from all base stations, Wi-Fi access points and mobile devices.The proposed system enables the logical network slicing into multiple isolated segments over a single physical infrastructure by using virtualization of radio resources. A new method of intelligent radio resource management is proposed to improve the efficiency of spectrum utilization by predicting traffic demand of particular network sliced using deep recurrent neural network. Simulation results show that for the case of network slicing the proposed method allows to double the capacity of the physical network infrastructure, while ensuring the target quality of service for all users.