I - WLAN : Intelligent Wireless Local Area Networking

I-WLAN: Intelligent Wireless Local Area Networking by Mustafa Ergen Doctor of Philosophy in Engineering-Electrical Engineering and Computer Sciences University of California, Berkeley Professor Pravin Varaiya, Chair Future wireless networks require intelligent components that can automate, scale and manage the network in order to handle the demand for ubiquitous access. In this dissertation, we first evaluate existing problems, and then we introduce components that improve the performance of existing systems. We introduce a new Markov model for Distributed Coordination of Function of IEEE 802.11 with which we formulate the throughput and delay for saturated and non-saturated traffic. We introduce a novel formulation for individual throughput when stations operate with mixed data rates. We introduce an admission control mechanism to maintain the highest achievable throughput by controlling the access. After that we use our throughput formulation to obtain the performance of an indoor network. We introduce a packet size adjustment scheme with respect to data rate so that a slow station sends small packets to prevent throttling of the network. We introduce a frame aggregation scheme for wireless 1 voice over IP. We introduce a fast and fair sub-optimal algorithm to allocate sub-carriers and bits adaptively in an OFDMA system for point-to-multipoint communication, and investigate MAC performance of the wireless LAN with adaptive antennas for point-topoint communication. We find that ignoring the consecutive transmission probability in the previous Markov models is incomplete. Consequently, our model which takes this into account is closer to the standard. We find that the individual throughput of a station is the same for fast and slow stations, and slow stations throttle the performance, since the CSMA/CA scheme gives equal access but not equal time of channel usage. Our packet size optimization scheme increases the throughput of the total network and the fast station but not that of the slow station. The admission control mechanism can tune the network from random access to controlled access. Network management can monitor the whole network and do real-time adjustments for optimum performance. Our frame aggregation scheme for wireless voice over IP reduces the number of access by concatenating the packets in the accesses point in order to be sent at one time. The results of our resource allocation scheme for adaptive sub-carrier and bit allocation is appealing and can be close to optimal. We find that a directional antenna increases the performance significantly and can provide enhancements for wireless LANs. Prof. Pravin Varaiya Dissertation Committee Chair