Optimal User Scheduling for Downlink Multi-User MIMO Systems Using Convolutional Neural Networks

Le résumé fourni par la source

Multi-user Generalized MIMO: Capacity and Performance Improvements, 99th European Wireless Conference, 2023. A practically crucial challenge for MIMO systems is the implementation of a user scheduling that selects only a subset of users for service at each time slot. These traditional user scheduling methods are either designed for a specific scenario or are time-consuming and unsuitable for highly dynamic systems, especially large-scale ones. The potential of artificial intelligence algorithms and intense learning models, such as convolutional neural networks (CNNs), to address sophisticated challenges across multiple domains, such as wireless communications systems, has gained momentum over the years. This work proposes a deep learningbased optimal user scheduling framework for the downlink multi-user MIMO systems. The training trains the network to predict which channel state information and system constraints users would choose to select the optimum subset. It also improved system throughput with quality of service constraints while evading global contention. However, the proposed method is fast, real-time, simple to implement and offers maximum gain over the prior approaches. Simulation results demonstrate that the proposed method can significantly outperform existing user scheduling methods of the same or even lower complexity, thus indicating the effectiveness and efficiency of multi-user MIMO user scheduling using CNN. Our deep learning—based approach enhances its performance and potential to cater to future wireless communication networks.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.