Slicing capacity-centered mode selection and resource optimization for network-assisted full-duplex cell-free distributed massive MIMO systems  

作　　者：Jie WANG Jiamin LI Pengcheng ZHU Dongming WANG Hongbiao ZHANG Yue HAO Bin SHENG 

机构地区：[1]National Mobile Communications Research Laboratory,Southeast University,Nanjing 210096,China [2]Purple Mountain Laboratories,Nanjing 211111,China [3]China Mobile Research Institute,Beijing 100032,China

出　　处：《Science China(Information Sciences)》2024年第1期199-216,共18页中国科学（信息科学）（英文版）

基　　金：supported in part by National Key Research and Development Program (Grant No.2021YFB2900300);National Natural Science Foundation of China (Grants Nos.61971127,61871122);Southeast University-China Mobile Research Institute Joint Innovation Center;Major Key Project of PCL (Grant No.PCL2021A01-2)。

摘　　要：Network-assisted full-duplex(NAFD) cell-free distributed massive multiple-input multipleoutput(MIMO) systems enable uplink(UL) and downlink(DL) communications within the same timefrequency resources,which potentially reduce latency by avoiding the overhead of switching UL/DL modes.However,how to choose UL/DL modes remains an important factor affecting system performance.With the dramatic increase in the number of users and access points(APs),massive access brings significant overhead in the mode selection.Additionally,the different quality of service(QoS) among users also makes the effective utilization of resources difficult.As one of the most promising technologies in sixth-generation(6G),network slicing enables the adaptive configuration of limited UL/DL resources through the resource isolation assisted NAFD technique.Therefore,we propose a slicing capacity-centered scheme.Under this scheme,APs are motivated by slicing requirements and associated slices to form different subsystems.Collaborative mode selection and resource allocation are performed within each subsystem to reduce overhead and improve resource utilization.To implement this scheme efficiently,a double-layer deep reinforcement learning(DRL)mechanism is used to realize the joint optimization of mode selection and resource allocation.Simulation results show that the slicing capacity-centered scheme can effectively improve resource utilization and reduce overhead.

关 键 词：network-assisted full-duplex network slicing mode selection resource optimization deep rein-forcement learning 

分 类 号：TN929.5[电子电信—通信与信息系统]