Robust Cross-Layer Design With Reinforcement Learning for IEEE 802.11n Link Adaptation
This paper proposes a link adaptation method for IEEE 802.11n, which can foresightedly co-optimize the Modulation and Coding Scheme (MCS) in the PHY layer and the frame size in the MAC layer. The link adaptation method employs Markov Decision Process (MDP) for modeling this cross-layer design. By solving the MDP model with a reinforcement learning which does not require a prior knowledge about the wireless environment, the foresighted transmission strategy can be computed. The simulation results verify the proposed method and show that their proposed method can improve the goodput by 25% at most, compared with the MCS-oriented link adaptation method.