Enhancing IEEE 802.11 Standard with Deep Reinforcement Learning for Optimal Channel Access

According to IEEE 802.11 standard, the collision avoidance mechanism is not the most efficient as it relies on a binary exponential backoff (BEB) algorithm. This algorithm increases the backoff interval whenever a collision is detected, aiming to reduce the likelihood of future collisions. The influence of this algorithm causes bandwidth wastage and decreases network performance dramatically in dense networks. Moreover, an incorrect backoff setting leads to more collision occurrences, making channel access very challenging. This work proposes using reinforcement learning (RL) algorithms, namely Deep Q Learning (DQN) and Deep Deterministic Policy Gradient (DDPG), to solve such an optimization problem. The proposed approach uses the average transmission queue level as an observation metric, with throughput as the reward and contention window value as the action. Simulation results based on NS3-gym show that DQN and DDPG outperform BEB for a dynamically increasing number of stations (dynamic scenario), showing a 45.52% increase in throughput with 50 stations.