Deep Learning Based Resource Allocation Method to Control System Capacity and Fairness for MU-MIMO THP

Abstract

This paper proposes a deep-learning-based resource allocation method to adaptively control system capacity and fairness for multi-user multiple-input and multiple-output (MU-MIMO). In the proposed method, Tomlinson-Harashima precoding (THP) is used to enhance the transmission rate. Additionally, channel resources are appropriately allocated based on user scheduling techniques, i.e., semiorthogonal user selection (SUS) for throughput maximization and proportional fairness (PF) for fairness among users. The primary feature of the proposed method is that it appropriately allocates channel resources by utilizing the user position information and target fairness index (FI) through deep learning. This makes it possible to meet various service requirements. Numerical simulations are used to demonstrate the effectiveness of the proposed method in terms of system capacity and fairness under different MIMO configurations and user distributions.