Demand driven adaptive transmission for wireless systems with heterogeneous traffic

With the increasing consumer demand for high-speed wireless access to the Internet, new technologies have been developed and standards have been proposed to improve the efficiency of wireless devices to provide the bandwidth and to support the QoS required. In this thesis, we investigate one aspect of the problem arose from providing Internet-like services with a wireless physical layer interface, specifically, energy optimal cross-layer designs of packet schedulers under strict, per-packet delay constraints. The design of the schedulers takes into consideration the randomness of the packet arrival as well as time-varying channels. Firstly, a novel convex optimization formulation is proposed. By assuming prescient knowledge of the channel state information and packet arrival and expiry times, an interesting analytical solution is derived with a novel geometric interpretation, referred to as piecewise water-filling. An efficient algorithm for calculating such a solution is also presented. The problem is considered under both single- and multi-carrier scenarios with simulation results showing improvements due to the channel diversity effect. In addition to the analysis of the prescient scheduler, practical issues are also considered. Specifically, several optimal causal schedulers are proposed, each assuming various degrees of prior knowledge of the system parameters. Through simulations of these causal schedulers, it was established that the optimal packet scheduler without cross-layer knowledge uses 10dB more energy than optimal prescient scheduler. In contrast, a practical causal scheduler with cross-layer knowledge of physical layer channel gains, utilizing an 8-tap Wiener channel prediction filter, can achieve energy usage that is only 3dB away from the prescient scheduler. For completeness, we also study the modifications required to the scheduling problem formulation for multi-user channels, specifically, under the information theoretical MAC and BC channel models. For practical systems, we propose a practical scheme for cross-layer design in a multi-user environment based on time division multiple access.