Theory of rational decision-making and its applications to adaptive transmission: Dissertation

In this thesis, adaptive transmission power control algorithms for reliable communication in channels with state are explored and further developed. In channels with state, strict adherence to Shannon-sense capacity may lead to very conservative system designs. In many practical systems, error-free communication is not required because these systems can cope with decoding errors. These considerations give rise to other information-theoretic notions where the rate of reliable communications is considered a random variable which depends not only on the statistical properties of the channel but also on the adaptive transmission strategy. Numerous studies on adaptive transmission in channels with state have already been conducted using expected value of communication rate or information outage probability as the relevant performance metrics. However, these metrics, although intuitively pleasing, have usually been introduced without rigorous justification. This thesis contributes to the state of the art in a number of ways. These include the development of new conceptual viewpoints on performance assessment of adaptive communication systems in channels with state as well as a new set of adaptive transmission power control algorithms. In particular, the models and methods of rational decision theory are introduced and systematically used in developing a unified framework for analysis and optimization of adaptive transmission in channels with state. The proposed framework properly addresses the limitation of finite coding length, takes into account the decision maker's preferences, considers uncertainties relevant in a given decision, and determines the optimal decision by maximizing some numerical index. A central finding of the theoretical studies is that many of the previously proposed performance metrics can be rigorously justified within the newly proposed framework. In addition, adaptive transmission power control in parallel Gaussian channels is considered with the aim of obtaining new classes of algorithms. The safety-first approach, risk theory, and expected utility theory are applied to derive novel transmission power control algorithms. The performance of the proposed power control algorithms is evaluated by computer simulations and compared against the performance of some other well-known algorithms.