Alocação de recursos em redes sem fio utilizando algoritmos baseados em aprendizagem de máquina
Nenhuma Miniatura disponível
Data
2024-02-01
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Federal de Goiás
Resumo
With the increasing rise in the number of mobile devices and intelligent IoT devices, wireless networks have become increasingly complex, autonomous, and heterogeneous in terms of the types of network architectures they incorporate. Within these networks, efficiently allocating and managing resources for users is a significant problem to be solved. In this context, deep reinforcement learning techniques are expected to be among the main technologies used to achieve global optimization in dynamic resource allocation. This paper presents a proposal for resource allocation in wireless networks, termed Cross-Entropy Reinforcement Learning, in order to maximize energy efficiency and meet users’ quality of service (QoS) requirements. The approach considers a multi-user, multi-objective communication system based on CP-OFDM (Cyclic-Prefix Orthogonal Frequency Division Multiplexing) technology using reinforcement learning methods,
specifically a Deep Q-Network (DQN) associated with the CrossEntropy algorithm to obtain an optimal resource allocation policy. The implementation encompasses system parameters such as bandwidth, modulation, number of users, and average packet size, while detailing the structure of resource elements, subcarrier distribution, and transmission capacity in different modulation modes within the LTE (Long Term Evolution) system frame. Simulation results demonstrate that the proposed method exhibits good convergence characteristics and performs better than the traditional DQN approach without cross-entropy.
Descrição
Palavras-chave
Alocação de recursos em redes sem fio, DQN adaptativa, Entropia cruzada, QoS, Resource allocation in wireless networks, Adaptative DQN, Cross-Entropy
Citação
LOPES, Adriano Ferreira; SILVA, Jean Lucas Barbosa. Alocação de recursos em redes sem fio utilizando algoritmos baseados em aprendizagem de máquina. 2024. 14 f. Trabalho de Conclusão de Curso (Bacharelado em Engenharia de Computação) - Escola de Engenharia Mecânica, Elétrica e de Computação, Universidade Federal de Goiás, Goiânia, 2024.