Overloaded multiuser communications: Analysis, signal design, and fair capacity

This thesis presents a general theory for the analysis and design of overloaded multiuser communication systems. In such systems, users simultaneously access a common channel whose total bandwidth is smaller than the aggregate of single-user bandwidths. Overloaded systems are thus important when bandwidth is at a premium, but the drawback is high interference. In this thesis, we focus on user-separation to mitigate this interference. It is usually achieved via signal design and power control. We contend, however, that power control can impose unrealistic demands on the users resources, and therefore we consider only signal design. In fact, our design benefits from the typically disparate (unconstrained) powers. In addition, our approach emphasizes fairness among users, and it seeks the right trade-off between receiver-complexity and performance.; This thesis is divided into two parts. The first part derives the asymptotic (low-noise limit) error performance of suboptimum receivers that have linear, group, and decision feed-back structures. In the process, we propose new receivers that are better suited to overloaded systems. We then design signals to optimize the asymptotic error performance. In this regard, we show that the linear receiver structure is severely limited (even with power control), but that with decision feedback, bandwidth-efficient signals can be systematically designed to meet user-specified Quality of Service (QoS) requirements.; The second part considers the information-theoretic limits on reliable communication. In particular, we design signals to optimize the capacity region using a QoS-based formulation. This approach is intrinsically fair. On the other hand, a single-letter characterization of the capacity region is desirable. In this regard, the well-known sum and symmetric capacity measures can be unfair, especially if powers are disparate. This motivates the definition of a finer, more iv equitable measure, namely the fair capacity. Our QoS-based signal design approach addresses the otherwise intractable problem of finding signals that maximize the fair capacity.