Topics In Modeling, Analysis And Optimisation Of Wireless Networks

Abstract

The work in this thesis is concerned with two complementary aspects of wireless networks research; performance analysis and resource optimization. The first part of the thesis focusses on the performance analysis of IEEE 802.11(e) wireless local area networks. We study the distributed coordination function (DCF) and the enhanced distributed channel access (EDCA) MAC of the IEEE 802.11(e) standard. We consider n IEEE 802.11(e) DCF (EDCA) nodes operating as a single cell; by single cell, we mean that every packet transmission can be heard by every other node. Packet loss is attributed only to simultaneous transmissions by the nodes (i.e., collisions). Using the well known decoupling approximation [19], we characterize the collision behaviour and the throughput performance of the WLAN with a set of fixed point equations involving the backoff parameters of the nodes. We observe that the fixed point equations can have multiple solutions, and in such cases, the system exhibits multistability and short-term unfairness of throughput. Also, the fixed point analysis fails to characterize the average system behaviour when the system has multiple solutions. We then obtain sufficient conditions (in terms of the backoff parameters of the nodes) under which the fixed point equations have a unique solution. For such cases, using simulations, we observe that the fixed point analysis predicts the long term time average throughput behaviour accurately. Then, using the fixed point analysis, we study throughput differentiation provided by the different backoff parameters, including minimum contention window (CWmin), persistence factor and arbitration interframe space (AIFS) of the IEEE 802.11e standard. Finally, we extend the above results to the case where the receiver supports physical layer capture. In the second part of the thesis, we study resource allocation and optimization problems for a variety of wireless network scenarios. For a dense wireless network, deployed over a small area and with a network average power constraint, we show that single cell operation (the channel supports only one successful transmission at any time) is throughput efficient in the asymptotic regime (in which the network average power is made large). We show that, for a realistic path loss model and a physical interference model (SINR based), the maximum aggregate bit rate among arbitrary transmitter-receiver pairs scales only as Θ(log(¯P)), where¯P is the network average power. Spatial reuse is ineffective and direct transmission between source destination pairs is the throughput optimal strategy. Then, operating the network with only a single successful transmission permitted at a time, and with CSMA being used to select the successful transmitter-receiver pair, we consider the situation in which there is stationary spatiotemporal channel fading. We study the optimal hop length (routing strategy) and power control (for a fading channel) that maximizes the network aggregate throughput for a given network power constraint. For a fixed transmission time scheme, we study the throughput maximizing schedule under homogeneous traffic and MAC assumptions. We also characterize the optimal operating point (hop length and power control) in terms of the network power constraint and the channel fade distribution. It is now well understood that in a multihop network, performance can be enhanced if, instead of just forwarding packets, the network nodes create output packets by judiciously combining their input packets, a strategy that is called “network coding.” For a two link slotted wireless network employing a network coding strategy and with fading channels, we study the optimal power control and optimal exploitation of network coding opportunities that minimizes the average power required to support a given arrival rate. We also study the optimal power-delay tradeoff for the network. Finally, we study a vehicular network problem, where vehicles are used as relays to transfer data between a pair of stationary source and destination nodes. The source node has a file to transfer to the destination node and we are interested in the delay minimizing schedule for the vehicular network. We characterize the average queueing delay (at the source node) and the average transit delay of the packets (at the relay vehicles) in terms of the vehicular speeds and their interarrival times, and study the asymptotically optimal tradeoff achievable between them.