DefinePK

Abstract

MmWave overlay 5G heterogeneous network (HetNet) proposed to boost the capacity of 5G networks has limited performance due to spectrum resource occupancy by backhaul links. Caching contents is a solution to solve the spectrum occupancy problem. Caching reduces the load on backhaul links and improves the utilization of access links. Most of the existing caching strategies use content popularity as the only factor and dont consider the priority of mobile nodes and need to provide differential QoS to users based on user characteristics and network dynamics. This work proposes a machine learning based strategy to manage the cache and backhaul resource allocation with the goal of providing differential QoS to user based on its priority and the current network dynamics at same time without degrading the average potential throughput. Cache management is adapted automatically to network dynamics, user characteristics and content characteristics using reinforcement learning.