Intelligent CSMA/CA for Wi-Fi networks
Résumé
Wireless communications have become essential to modern life. They allow user to stay connected to their applications while supporting their mobility requirements. Wi-Fi equipped devices are getting increasingly widespread, and high-bandwidth applications like video streaming are becoming more critical in many areas. Thus, many performance challenges and issues have emerged in the past decades. Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) is a Wi-Fi multiple access control method used to manage access to the channel and to mitigate collisions. However, it suffers from performance degradation when the density of Wi-Fi nodes or the offered load increase. The random behavior of CSMA/CA is one of the reasons behind the performance degradation. In this paper, we propose a Deep Reinforcement Learning (DRL) mechanism, Intelligent CSMA/CA (ICSMA/CA), to dynamically adapt the backoff duration of CSMA/CA algorithm in dense Wi-Fi environments. We train and evaluate our DRL model using the Network Simulator (NS-3) and Tensorflow. Results show an improvement in the capacity utilization of the channel and a reduction in the channel access delay.