Abstract
5G definition and standardization projects are well underway, and governing characteristics and major challenges have been identified. A critical network element impacting the potential performance of 5G networks is the backhaul, which is expected to expand in length and breadth to cater to the exponential growth of small cells while offering high throughput in the order of Gbps and less than one-millisecond latency with high resilience and energy efficiency. Such performance may only be possible with direct optical fibre connections which are often not available countrywide and are cumbersome and expensive to deploy. On the other hand, a prime 5G characteristic is diversity, which describes the radio access network, the backhaul, and also the types of user applications and devices. Thus, we propose a novel, distributed, selfoptimized, end-to-end user-cell-backhaul association scheme that intelligently associates users with potential cells based on corresponding dynamic radio and backhaul conditions while abiding by users’ requirements. Radio cells broadcast multiple bias factors, each reflecting a dynamic performance indicator (DPI) of the endto-end network performance such as capacity, latency, resilience, energy consumption, etc. A given user would employ these factors to derive a user-centric cell ranking that motivates it to select the cell with radio and backhaul performance that conforms to the user requirements. Reinforcement learning is used at the radio cell to optimize the bias factors for each DPI in a way that maximizes the system throughput while minimizing the gap between the users’ achievable and required end-to-end quality of experience (QoE). Preliminary results show considerable improvement in users QoE and cumulative system throughput when compared to state-of-theart user-cell association schemes.