Department of Electrical and Computer Engineering
Ph.D. Thesis Abstract
SERVICE OUTAGE BASED ADAPTIVE TRANSMISSION IN FADING CHANNELS
The service outage based allocation problem explores variable rate transmission schemes and combines the concepts of ergodic capacity and outage capacity for fading channels. The ergodic capacity determines the maximum achievable rate for non real-time applications, and the outage capacity is developed for constant rate real-time applications. Neither is completely appropriate for variable rate multimedia applications for next generation wireless networks. In this context, the service outage based allocation problem is proposed. A service outage occurs when the instantaneous transmission rate is smaller than a basic rate specied by an application. The service outage allocation problem is to nd the optimum power allocation that maximizes the average rate subject to a service outage probability constraint and an average power constraint. The optimum power allocation is derived for a single at fading channel, and is generalized to M parallel fading channels. Two near optimum schemes are also derived for a small outage probability. The minimum outage based near optimum scheme significantly reduces the computational complexity. The allocation problem with respect to the energy efficiency is also examined for the M = 1 fading channel. In the application of the transmission of mixed real-time and non real-time services in fading channels, the optimum service outage based allocation can be implemented using an adaptive channel partition approach. The optimum fixed channel partition scheme is derived and compared to the adaptive scheme. In addition, a suboptimum fixed partition scheme called proportional average power partition is studied and observed to be close to the optimum fixed partition scheme in the Rayleigh fading channel.
We also study the performance of variable-rate turbo bit-interleaved coded modulation (Turbo-BICM) with random puncturing. A union-Bhattacharyya rate threshold for the Turbo-BICM based on a reliable channel region for turbo codes transmitted over parallel-channel is derived. A closed form approximation of this rate threshold is determined for an AWGN channel. This rate threshold is shown to predict the Turbo- BICM iterative decoding performance very well. Adaptation of Turbo-BICM in a slow fading channel is studied. The optimum power and modulation allocations are described. A dual problem solution which achieves a rate close to the optimum solution with significantly reduced computational complexity is described. Two simple schemes: water-filling with optimum modulation and equal power allocation with optimum modulation are also presented and shown to achieve a good performance. Proposed adaptive schemes are shown to achieve a rate within 2-3 dB of the ergodic capacity of a Rayleigh fading channel.
Ph.D. Dissertation Director: Professor Roy D. Yates
Download the Thesis: PDF