Degrees of freedom of wireless interference network
Zhengdao Wang,Lei Ke
DOI: https://doi.org/10.31274/etd-180810-2296
2011-01-01
Abstract:Wireless communication systems are different from the wired systems mainly in three aspects: fading, broadcast, and superposition. Wireless communication networks, and multi-user communication networks in general, have not been well understood from the information-theoretic perspective: the capacity limits of many multi-user networks are not known. For example, the capacity region of a two-user single-antenna interference channel is still not known, though recent result can bound the region up to a constant value. Characterizing the capacity limits of multi-user multiple-input multiple-output (MIMO) interference network is usually even more difficult than the single antenna setup. To alleviate the difficulty in studying such networks, the concept of degrees of freedom (DoF) has been adopted, which captures the first order behavior of the capacities or capacity regions. One important technique developed recently for quantifying the DoF of multi-user networks is the so-called interference alignment. The purpose of interference alignment is to design the transmit signals structurally so that the interference signals from multiple interferers are aligned to reduce the signal dimensions occupied by interference. In this thesis, we mainly study two problems related to DoF and interference alignment: (1) DoF region of MIMO full interference channel (FIC) and Z interference channel (ZIC) with reconfigurable antennas, and (2) the DoF region of an interference network with general message demands. For the first problem, we derive the outer bound on the DoF region and show that it is achievable via time-sharing or beamforming except for one special case. As to this particular special case, we develop a systematic way of constructing the DoF-achieving nulling and beamforming matrices. Our results reveal the potential benefit of using the reconfigurable antenna in MIMO FIC and ZIC. In addition, the achievability scheme has an interesting space-frequency interpretation. For the second problem, we derive the DoF region of a single antenna interference network with general message demands, which includes the multiple unicasts and multiple multicasts as special cases. We perform interference alignment using multiple base vectors and align the interference at each receiver to its largest interferer. Furthermore, we show that the DoF region is determined by a subset of receivers, and the DoF region can be achieved by considering a smaller number of interference alignment constraints so as to reduce the number of time expansion. Finally, as a related research topic, we also include a result on the average throughput of a MIMO interference channel with single-user detector at receivers and without channel state information at transmitters. We present a piecewise linear approximation of the channel throughput under weak, moderate and strong interference regimes. Based on that we determine the optimal number of streams that a transmitter should use for different interference levels.