Luo HHUNG-YU WEI2023-06-092023-06-09202115502252https://www.scopus.com/inward/record.uri?eid=2-s2.0-85123024758&doi=10.1109%2fVTC2021-Fall52928.2021.9625307&partnerID=40&md5=1320416a5d6e879ce737d54df36f2026https://scholars.lib.ntu.edu.tw/handle/123456789/632476Millimeter wave (mmWave) is a crucial component in 5G and beyond 5G communications. However, the dense deployment of mmWave transceivers imposes a heavy burden on the management of radio access network (RAN). This challenge increases the need for autonomous network management methods leveraging machine learning (ML) techniques. In particular, mmWave beam selection is a critical issue for the management of RAN due to the large training overhead on mmWave transceivers. To this end, a new beam tracking method based on sequence-to-sequence (Seq2Seq) learning is proposed. Besides, thanks to edge computing technologies, network management algorithms and delay-sensitive user applications can be hosted on edge servers in close proximity. Due to limited resources on the edge server, the resource allocation problem for beam tracking and edge gaming is investigated with the aim of maximizing game quality of experience (QoE). Simulation results verify the effectiveness of the proposed orchestration scheme. © 2021 IEEE.edge computing; machine learning; mmWave communication; network management; resource allocation[SDGs]SDG95G mobile communication systems; Delay-sensitive applications; Edge computing; Machine learning; Network management; Quality of service; Radio access networks; Radio transceivers; Autonomous networks; Edge computing; Edge server; Machine-learning; Management method; Millimeter wave transceivers; Millimeterwave communications; Networks management; Radio access networks; Resources allocation; Resource allocationconference paper10.1109/VTC2021-Fall52928.2021.96253072-s2.0-85123024758