Framework for Slice-Aware Radio Resource Management Utilizing Artificial Neural Networks

Abstract

For accommodating the heterogeneous services that are anticipated for the fifth-generation (5G) mobile networks, the concept of network slicing serves as a key technology. Spanning both the core network (CN) and radio access network (RAN), slices are end-to-end virtual networks that share the resources of a physical network. Slicing the RAN can be more challenging than slicing the CN since RAN slicing deals with the distribution of radio resources, which have fluctuating capacity and are harder to extend. Improving multiplexing gains, while assuring the slice isolation is the main challenging task for RAN slicing. This paper provides a flexible and configurable framework for RAN slicing, where diverse requirements of slices are simultaneously taken into account, and slice management algorithms adjust the control parameters of different radio resource management (RRM) mechanisms to satisfy the slices' service level agreements (SLAs). One of the proposed algorithms is based merely on heuristics and the other one utilizes an artificial neural network (ANN) to predict the behavior of the cellular network and make better decisions in the adjustment of the RRM mechanisms. Furthermore, a protection mechanism is devised to prevent the slices from negatively influencing each other's performances. A simulation-based analysis demonstrates that in presence of local or global overload of one of the slices, the ANN-based method increases the number of key performance indicators (KPIs) that fulfill their defined SLA targets. Finally, we show that the proposed protection mechanism can force the negative effects of an overloading slice to be contained to that slice and the other slices are not affected as severely.

Keywords

• Network slicing
• radio resource management
• slice orchestration
• 5G
• iterative adaptation
• artificial neural networks