Cross-layer optimization for transmission of delay- sensitive and bursty traffic in wireless systems

High demands on the quality of service (QoS), in terms of throughput, delay, and packet loss, in wireless systems have fueled substantial research interest in jointly considering different layers of protocol stack. Integrated design of wireless systems from physical to application layers is challenging due to high variability of wireless channels and stochastic and delay-sensitive nature of traffic. Following this research interest, the current dissertation considers resource allocation in wireless systems for transmission of delay-sensitive and bursty traffic in two main areas. In the first chapter of this dissertation, a subcarrier allocation problem in OFDMA downlink system is studied, where given the knowledge of the channel and queue states, the optimal centralized allocation policy seeks to minimize the average packet delay. The problem is modeled as a multi-queue multi- server assignment problem with time-varying connectivities. For on-off connectivities and homogeneous users, we show, using a dynamic programming approach, that a simultaneous maximum-throughput and load-balancing policy is delay-optimal. For more general connectivities, we propose heuristic policies that use different degrees of queue and channel state information to provide good delay performance for various traffic loads. The rest of the dissertation is concerned with cross-layer design of wireless data networks, when there is no channel state information at the transmitter and no retransmission. In Chapter 3, we study how to set up various physical layer parameters, e.g., coding block length and channel transmission rate of point-to-point wireless fading channels, such that the total probability of bit loss is minimized, where bit losses account for both erroneous decoding at the receiver as well as violation of a specified delay constraint. We simplify the problem by considering an asymptotic high signal-to-noise-ratio (SNR) regime and assuming smoothly scaling (with SNR) bit- arrival processes. Extending this study to multi-user settings, in Chapter 4 we study how to select the optimal channel spatial-diversity in MIMO multiple access channels in order to minimize the asymptotic high-SNR error probability. We also quantify the amount of the performance improvement that can be achieved from using an optimal queue-aware dynamic rate scheduler. While Chapter 4 answers the above question for sufficiently large delay constraints, Chapter 5 considers a case of finite and small buffer constraints. Finally, in Chapter 5, we propose a large-deviations analysis of the asymptotic buffer overflow probability for a maximum-weight dynamic scheduling policy with simplex rate region, assuming properly-scaled arrival processe