Principles of proactive resource allocation in wireless communication networks

The excessively growing demand on the wireless data services has raised major concerns of a potential degradation, if not a total collapse, of satisfactory mobile communications. On the other hand, the available spectrum for wireless communications has been reported to suffer from a daily underutilization problem that lasts from midnight to early morning hours. Such a discrepancy between the wireless traffic levels over the course of the day is essentially tied to the human activity patterns, whereby end users exhibit high demand characteristics during the day time creating the so-called peak hour load, and concurrently idle at the late night time yielding substantially low demand and the so-called off-peak hour load.;Major research efforts have been exerted over the past few years to develop a radical remedy to such a problem threatening the future of high-quality wireless communications. However, almost all of the emerging solutions, including cognitive radio communications, time-dependent pricing, and WiFi offloading, rely on influencing the economical responsiveness of wireless users to delay their demand from the peak to the off-peak time. The resulting gains of these proposed solutions hinge on the tradeoff between the offered pricing incentives and the flexibility of the users to change their activity patterns.;In this dissertation we bring to attention an unexploited degree of freedom in the realm of wireless resource allocation. That is, the human behavior is highly predictable. Motivated by the recent findings that affirm this observation, we propose the proactive resource allocation paradigm in wireless networks. In particular, we investigate the design of optimal proactive data download policies that service predictable peak hour demand during the off-peak time without affecting the activity patterns of end users, and study their potential gains. Starting with the idealistic scenario of fully predictable requests and essentially static data content, we show that the predictability of users' demand T slots in advance yields an exponentially improved outage/blocking probability in a capacity limited scenario, with an exponent that grows at least linearly in T..;Further, we study the realistic scenario of inaccurate predictions with dynamic content and we show that service provider can employ proactive download policies that attain a cost reduction that unboundedly grows with the number of data items and number of users. Finally, we consider the economical aspects of proactive resource allocation and design joint smart data pricing and proactive content download policies that (1) boost the profit of the service providers, (2) reduce the expected payments by the end-users, and (3) maintain end-users activity characteristics unchanged over time. These potential win-win situations are driven by the cost reduction promised by proactive data services, as well as corresponding pricing incentives to render the users' demand more deterministic.