Learning Based Resource Allocation Under Limited Backhaul Capacity of Millimeter Wave Networks

Abstract

Millimeter wave (mmwave) communication, which is an ideal candidate for meeting the high data rate requirements, are prone to random and dynamic blockages due to the small wavelength used. Backhaul communication between base station and core network, on the other hand, provides a stable connection, but have limited capacity when compared with increasing data rate requirements at the user side. Efficient backhaul resource allocation is required to strike a balance between the limited capacity, stable backhaul link and high capacity, unstable mmwave links. We propose a combination of stochastic geometry and analytical channel model to bring in real world mmwave blockage effect. A neural network based backhaul resource allocation is proposed to maximize the system utility under limited backhaul capacity of a mmwave network. Simulation results prove that the proposed method outperforms the existing solutions.