Low-overhead cooperation to mitigate interference in wireless networks

Wireless cellular networks, which serve a large area by geographically partitioning users, suffer from interference from adjacent cells transmitting in the same frequency band. This interference can theoretically be completely mitigated via transceiver cooperation in both the uplink and downlink. Optimally, base stations serving the users can utilize high-capacity backbones. to jointly transmit and receive all the data in the network across all the base stations. In reality, the backbone connecting the base stations is of finite capacity, limiting joint processing to localized clusters. Even with joint processing on a small scale, the overhead involved in sharing data between multiple base stations is large and time-sensitive. Other forms of cooperation have been shown to require less overhead while exhibiting much of the performance benefit from interference mitigation. One particular strategy, called interference alignment (IA), has been shown to exploit all the spatial degrees of freedom in the channel provided data cannot be shared among base stations. Interference alignment was developed for the multi-user interference channel to exploit independent channel observations when all of the links in the network have high signal-to-noise ratio, and assumes all the nodes utilizing the physical resources are participating in the cooperative protocol. When some or all of the links are at moderate signal-to-noise ratio, or when there are non-cooperating users, IA is suboptimal. In this dissertation, I take three approaches to addressing the drawbacks of IA. First, I develop cooperative transmission strategies that outperform IA in various operationg regimes, including at low-to-moderate SNR and in the presence of non-cooperating users. These strategies have the same complexity and overhead as IA. I then develop algorithms for network partitioning by directly considering the overhead of cooperative strategies. Partitioning balances the capacity gains of cooperation with the overhead required to achieve them. Finally, I develop the shared relaying model, which is equivalent to the interference channel but with a single multi-antenna relay mediating communications between transceivers. The shared relay requires less overhead and cooperation than interference alignment but requires added infrastructure. It is shown to outperform conventional relaying strategies in cellular networks with a fixed number of total relay antennas.