This paper details optimizing the performance of Wi-Fi networks under Double Deep Q-Network (DDQN) Learning. We tackle aspects of optimization in term of adapting the throughput depending on the state of the network in order to avoid overloading the network one technique for achieving these adaptations is the employment of machine learning. In this paper, we employ Double Deep Q-Network (DDQN) learning, a branch of machine learning, to optimize the performance of Wi-Fi network. We explore the potential of DDQN learning in enhancing application profile selection within congested Wi-Fi networks, where performance can deteriorate significantly due to saturation. We develop and assess a DRL mechanism to identify the most suitable application profiles for achieving optimal network performance under any given network condition. We evaluated the produced DDQN through simulation. Results showed that the proposed mechanism achieved better throughput performance during network saturation compared to a similar approach.
Farhad Bahadori-Jahromi (Thu,) studied this question.