Performance Optimization in Wireless Local Area Networks

Wireless Local Area Networks (WLAN) are becoming more and more important for providing wireless broadband access. Applications and networking scenarios evolve continuously and in an unpredictable way, attracting the attention of academic institutions, research centers and industry. For designing an e cient WLAN is necessary to carefully plan coverage and to optimize the network design parameters, such as AP locations, channel assignment, power allocation, MAC protocol, routing algorithm, etc... In this thesis we approach performance optimization in WLAN at di erent layer of the OSI model. Our rst approach is at Network layer. Starting from a Hybrid System modeling the ow of tra c in the network, we propose a Hybrid Linear Varying Parameter algorithm for identifying the link quality that could be used as metric in routing algorithms. Go down to Data Link, it is well known that CSMA (Carrier Sense Multiple Access) protocols exhibit very poor performance in case of multi-hop transmissions, because of inter-link interference due to imperfect carrier sensing. We propose two novel algorithms, that are combining Time Division Multiple Access for grouping contending nodes in non-interfering sets with Carrier Sense Multiple Access for managing the channel access behind a set. In the rst solution, a game theoretical study of intra slot contention is introduced, in the second solution we apply an optimization algorithm to nd the optimal degree between contention and scheduling. Both the presented solutions improve the network performance with respect to CSMA and TDMA algorithms. Finally we analyze the network performance at Physical Layer. In case of WLAN, we can only use three orthogonal channels in an unlicensed spectrum, so the frequency assignments should be subject to frequent adjustments, according to the time-varying amount of interference which is not under the control of the provider. This problem make necessary the introduction of an automatic network planning solution, since a network administrator cannot continuously monitor and correct the interference conditions su ered in the network. We propose a novel protocol based on a distributed machine learning mechanism in which the nodes choose, automatically and autonomously in each time slot, the optimal channel for transmitting through a weighted combination of protocols.