With the growing number of users along with ever-increasing demand for higher data rates and better quality of service in modern wireless networks, interference has become the major barrier against efficient utilization of limited resources. On the other hand, opportunities for cooperation among radios also increase with the growing number of users, which potentially lead to better interference management. In traditional wireless system design, however, such opportunities are usually neglected and only basic interference management schemes are employed, mainly due to lack of fundamental understanding of interference and cooperation.

In this dissertation, we study the fundamental aspects of cooperative interference management through the lens of network information theory. In the first and the second part, we characterize both qualitatively and quantitatively how limited cooperation between transmitting or receiving terminals helps mitigate interference in a canonical two-transmitter-two-receiver wireless system. We identify two regions regarding the gain from limited cooperation: linear and saturation regions. In the linear region cooperation is efficient and provides a degrees-of-freedom gain, which is either one cooperation bit buys one bit over the air or two cooperation bits buy one bit over the air until saturation. In the saturation region cooperation is inefficient and only provides a bounded power gain. The conclusions are drawn from the approximate characterization of the capacity regions.

In the third part, we investigate how intermediate relay nodes help resolve interference in delivering information from two sources to their respective destinations in multi-hop wireless networks. We focus on a linear deterministic approximate model for wireless networks, and when the minimum cut value between each source-destination pair is constrained to be 1, we completely characterize the capacity region. One of the interesting findings is that, at most four nodes need to take special coding operations so that interference can be canceled over-the-air or within-a-node, while other nodes can take oblivious operations. We also develop a systematic approach to identify these special nodes.