Investigation on Offset Optimization for Cell Range Expansion Using Neural Networks in Conjunction with Convex Optimization

Abstract

In order to support increasing amount of traffic, network densification, and additional higher frequency usage is one of the promising techniques. To make efficient utilization of these increasing number of base stations (BSs) and frequency bands, radio resource allocation is important. In this paper, two types of resource allocation are used jointly: user association (UA), in which users select resources with channel quality information (CQI) feedback, and scheduling, in which actual allocation is decided based on CQI. UA was performed using cell range expansion (CRE), and offset optimization was performed using a neural network (NN). Convex optimization was employed for scheduling, and the two different types of resource allocation were integrated using the same proportional fair criteria to accomplish overall optimization. In this paper, we focus specifically on the training of NN. Simulation results show that the proposed method works well.