An Optimal Transmission Strategy for IEEE 802.11 Wireless LANs: Stochastic Control Approach

The family of IEEE 802.11 Wireless Local Area Network (WLAN) standards supports multiple transmission rates in the physical layer (PHY). This multi-rate capability offers a viable means of coping with dynamically-fluctuating wireless channel conditions. We propose a cross-layer approach for the optimal PHY mode control to maximize the system goodput. Our key idea is to exploit the underlying channel fading characteristics and the history of PHY mode control and observations to infer the current channel condition, so that the optimal PHY mode may be selected. Assuming Rayleigh fading, we describe the receiver-side signal-to-noise ratio (SNR) fluctuations as a finite-state Markov channel (FSMC) model. Since the channel condition (i.e., fading level) is not directly observable by the transmitter, we formulate the PHY rate adaptation problem as a partially-observable Markov decision process (POMDP) to find the optimal transmission policy. We use the belief state vector to represent the channel state probabilistically. The belief state is updated solely based on the channel state-transition matrix and acknowledgement (ACK) information. Our evaluation results show that the POMDP-based rate adaptation outperforms two most well-known rate adaptation schemes, Auto Rate Fallback (ARF) and Adaptive ARF (AARF), in terms of average goodput under various fading conditions, and achieves up to 92% of the ideal performance.