Multiaccess and fading in communication networks

Two fundamental issues in the design of wireless communication networks are the interference among multiple users and the time-varying nature of the fading wireless channel. We apply fundamental techniques in information theory and queueing theory to gain insights into the structure of these problems. In a terrestrial cellular or space network, multi-user interference arises naturally as different users in the same cell or region attempt to transmit to the base station or satellite at the same time and in the same frequency range. We first examine the impact of this interference on the design of error correction codes for reliable data transmission. At the physical layer of the wireless network, the phenomenon of multi-user interference is captured by the multiaccess (many-to-one) channel model. The set of all data rates at which reliable communication can take place over this channel is characterized via information theory by the so-called multiaccess capacity region. A basic problem is developing coding schemes of relatively low complexity to achieve all rates in this capacity region. By exploiting the underlying geometrical structure of the capacity region, we develop a method of reducing the multi-user coding problem to a set of single-user coding problems using the ideas of timesharing and successive decoding. Next, we investigate the effect of multi-user interference on higher-layer quality-of-service issues such as packet delay. Under certain conditions of symmetry, we find that the structure of the multiaccess capacity region can again be used to obtain a “load-balancing” queue control strategy which minimizes average packet delay for Poisson data sources. Due to the mobility of users and constantly changing multipath environments, wireless channels are inherently time-varying, or fading. Any sensible design of wireless networks must take into account the nature of this fading and the ability of the system to track channel variations. We consider a wireless system in which a single user sends time-sensitive data over a slowly varying channel. Information regarding the state of the channel is fed back with some delay to the transmitter, while the receiver decodes messages within some fixed and finite amount of time. Under these conditions, we demonstrate a provably optimal transmission strategy which maximizes the average data rate reliably sent across the wireless channel. The strategy is based on the information-theoretic idea of “successive refinement,” whereby the decoder decodes at different rates according to the observed channel state. Thesis Supervisor: Robert G. Gallager Title: Professor For Mother and Father