A utility-based resource allocation scheme for IEEE 802.11 WLANs via a machine-learning approach

The problem of allocating resources in IEEE 802.11 wireless local area networks (WLAN) is challenging due to limited bandwidth, time-varying channel conditions, and especially the distributed channel-access manner. In this paper we propose an intelligent resource allocation scheme that dynamically adjusts medium-access-control (MAC) parameters to tune channel-access opportunities and maximize the total utility. “Intelligent” refers to the capability of our approach to regulate each 802.11 node’s parameters automatically according to the changes of surrounding situations, e.g. channel conditions and number of nodes. Our intelligent allocation scheme uses neural networks to on-line learn the nonlinear function between the adopted MAC parameters and allocated throughput. Based on the learned knowledge, MAC parameters can therefore be dynamically adjusted toward the desired throughput allocation and consequently the maximal WLAN utility. Simulations results demonstrate the effectiveness of our allocation scheme in maximizing the system utility in a varying 802.11 WLAN environment.