ThesisInterference Management in Wireless Networks with Channel Uncertainty
Having multiple terminals access simultaneously the wireless medium enhances the capacity of cellular networks, whereas inevitably introduces multiuser interference, which in turn limits spectral efficiency. Hence, interference management constitutes a major means to improve spectral efficiency. Crucially, the spectral efficiency improvement by most interference management techniques usually stems from the assumption of the availability of channel state information at the transmitters (CSIT) obtained through feedback. The inaccuracy and/or latency of channel knowledge via feedback are major sources of channel uncertainty that affect network performance to a great extent. This is particularly true of methods relying on multiple-antenna precoding.
This thesis focuses on interference management with channel uncertainty in multiuser multiple-input multiple-output (MIMO) networks, where the channel uncertainty comes from feedback delays as well as the strictly limited capacity of feedback links. As such, in the extreme cases, the transmitters may either possess sufficiently precise CSI but with large latency or have access instantaneously to a coarse channel information (e.g., topological feedback, with one bit indicating whether the channel is strong or weak). The former renders the available CSI feedback obsolete under a fast fading channel, and the latter makes the transmitters almost blind to the exception of the binary indicator of channel strength. This thesis focuses on different regimes of CSIT availability, trying to address two fundamental problems: (1) How to best exploit delayed feedback? (2) How to best exploit drastically reduced feedback (e.g., such a topology-related feedback)?
In this thesis, above questions are addressed mainly from the information theoretic perspective, but also from signal processing and communication theoretic ones, from which some new interference management techniques are proposed to combat channel uncertainty and improve spectral efficiency, whereas the fundamental limitations of some existing interference management techniques are also revealed.