Metamodeling of Cross-layer Cooperative Scheduling in Wireless Local Area Networks

Multihop, ad-hoc Wireless Local Area Networks (WLAN) based on 802.11 are expected to serve as important segments of next generation wireless networks. In the network layer of intermediate nodes of such networks, a priority scheduler arbitrates the transmission between forwarding traffic and the node’s own locally produced traffic. Meanwhile, a new QoS-aware Enhanced Distributed Channel Access (EDCA) of 802.11e provides differentiated access at the MAC layer. Therefore, how to control priority forwarding in a cross-layer fashion, namely, by combining scheduling at the network layer with EDCA differentiated access at the MAC layer in order to achieve the best channel utilization is an important design issue. Due to the numerous control variables potentially involved from both layers, analytical modeling is not feasible. Instead of using a simulation model to predict some points of the utilization response surface, we propose to utilize a metamodel (used more frequently in other disciplines such as operations research) in order to approximate the complete response surface by fitting regression analysis over a large amount of simulation results. The resulting metamodel allows us to establish and quantify the cross-layer effects through specific interaction items and can serve for the purpose of cross-layer optimization.