Autonomous Traffic Engineering using Deep Reinforcement Learning

• Autor(a) principal: Lotfi, Afshin Nakhost
• Data de Publicação: 2022
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10362/155377
• topic: Traffic Engineering; Deep Reinforcement Learning; Load Balancing; Domínio/Área Científica::Engenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática

The evolution of communication technologies in the past few decades, has led to a huge increase in the complexity and the overall size of telecommunication networks. This phenomenon has increased the need for innovation in the field of Traffic Engineering (TE), as the already existing solutions are not flexible enough to adapt to these changes. With the appearance of 5G technologies, the urgency to revolutionize the field is higher than ever and the softwarization and virtualization of the infrastructure bring new possibilities for TE optimization, namely the possible use of Artificial Intelligence (AI) based methods for Traffic Management. The recent advances in AI have provided model-free optimization methods with algorithms like Deep Reinforcement Learning (DRL) that can be used to optimize traffic distributions in complex and hard to model Network scenarios. This thesis aims to provide a DRL-based solution for TE where an agent is capable of making routing decisions based on the current state of the network, with the goal of balancing the load between the network paths. A DRL agent is developed and trained in two different scenarios where the traffic that already exists in the network is generated randomly or according to a systematic pattern. A simulation environment was developed to train and evaluate the DRL agent.