Banca de DEFESA: DANIEL RODRIGUES DE LUNA
Uma banca de DEFESA de DOUTORADO foi cadastrada pelo programa.STUDENT : DANIEL RODRIGUES DE LUNA
DATE: 13/02/2023
TIME: 09:00
LOCAL: Remota via Google Meet (https://meet.google.com/oyd-stzu-azz)
TITLE:
Adaptive Bandwidth Partitioning in 5G NR Systems Using Machine Learning Solution
KEY WORDS:
PAGES: 135
BIG AREA: Engenharias
AREA: Engenharia Elétrica
SUBÁREA: Telecomunicações
SPECIALTY: Sistemas de Telecomunicações
SUMMARY:
The fifth generation (5G NR) of the 3GPP communication network proposes a variety of use cases, ranging from improved mobile broadband communications (eMBB) to ultra-reliable low latency communications (URLLC), in addition to massive communications between machines (mMTC). Introducing flexibility in bandwidth utilization is one of the critical requirements of 5G technologies. As such, the concept of bandwidth parts is introduced to give flexibility, fast-changing configurations, and energy saving to UEs that do not need the whole bandwidth available by using just a part of it. The use case that can benefit from this feature is mMTC, characterized by its massive number of devices and the need for small power consumption. This work proposes a reinforcement learning solution of bandwidth part adaptation in mMTC scenarios to save energy and improve system statistics. Firstly, the subject and a complete section of related works with the more recent papers are introduced, highlighting the gap in this area of research. In addition, a brief theoretical approach to 5G is presented as part of the basis of this work. Then, the system model and main system parameters are described, along with the simulation tool used, the ns-3 NR 5G-LENA, in which calibration campaigns are carried out to validate its use. Afterward, we detail the investigation scenario in which we can apply the reinforcement learning (RL) solution. After decentralized and centralized solutions are proposed, new campaigns are created using both proposed RL-based bandwidth part adaptation solutions. The final results campaign shows the gains attained compared to a traditional fixed approach. Finally, the papers published and the main discussions of this thesis are gathered in the end.
COMMITTEE MEMBERS:
Presidente - 1412682 - VICENTE ANGELO DE SOUSA JUNIOR
Interno - 1837240 - MARCELO AUGUSTO COSTA FERNANDES
Externo ao Programa - 1840342 - MARCIO EDUARDO DA COSTA RODRIGUES - UFRNExterno à Instituição - ANDRÉ MENDES CAVALCANTE
Externo à Instituição - WALTER DA CRUZ FREITAS JÚNIOR - UFC