The main driving force for the evolution of wireless systems is to provide increased data-rate. Over the past few years, many new wireless techniques have emerged in multiple communication layers to provide more efficient use of bandwidth, such as rateless codes, hybrid ARQ, multi-user MIMO, full-duplex transmission. In order to support the ever-increasing wireless traffic demand, it is crucial to understand the benefits of these schemes in terms of throughput gains, and develop efficient algorithms to reap the full potential of these new wireless techniques. Hence, in this dissertation, we focus on the characterization of throughput and the design of control algorithms in wireless system.
We first consider the point-to-point channel and focus on the distribution of delay, a metric that is closely related to link-throughput and determined jointly by the selection of physical-layer code-rate and link-layer retransmission scheme. Our investigations reveal the surprising results that, when decoder memory is not used to cache undecodable transmissions and there is a lack of redundancy in the packet, a light-tailed packet-size distribution may translate into a heavy-tailed transmission delay, leading to a poor link-throughput (possibly even zero-throughput), whereas the delay will be lighted-tailed if either the decoder uses memory to cache failed transmissions or the amount of redundancy in the packets passes a certain threshold. Next, we shift our focus from the point-to-point channel to the point-to-multipoint channel and investigate how the broadcast-throughput behaves as a function of network size and coding block size, when an optimal rateless codes is used. Using large deviation theory, we are able to obtain a close-form expression of the asymptotic throughput (asymptotic in the number of multicast receivers) for any mapping of the network size to the coding block size. This asymptotic throughput result leads us to find a lower-bound on the throughput for any finite values of coding block size and network size, which is also asymptotically tight.
We next shift our focus from link-layer throughput to the MAC-layer throughput-region in multi-hop wireless networks. By assuming a binary interference relationship between links, we provide a thorough comparison of the throughput-regions that can be achieved under different combinations of multi-antenna techniques, such as MIMO multiplexing, multi-user MIMO and wireless full-duplex, from a degree-of-freedom perspective. Our results give clear guidelines on which multi-antenna architecture and traffic pattern could result in throughput improvement for one scheme over another. While the throughput-regions of different schemes may be compared, the adoption of new schemes raises challenges in the design of control algorithms that aim to support the entire throughput-region. Indeed, when each node is capable of wireless full-duplex cut-through transmission, the MAC-layer throughput-region is directly a function of the routing decision, leading to a strong coupling between routing and scheduling, which has not been seen in the traditional half-duplex network. Also, it is unclear how to dynamically form/change cut-through routes based on the traffic rates and patterns. In this dissertation, we introduce a novel method to characterize the interference relationship between links in the network with cut-through transmission, which decouples the routing decision with the scheduling decision and enables a seamless adaptation of traditional half-duplex algorithms into wireless networks with full-duplex cut-through capabilities.