Interference alignment in MIMO interference channels with limited channel state information

다국어 초록 (Multilingual Abstract)

In a wireless network with multiple interfering links, interference alignment achieves the maximum degrees-of-freedom by aligning all interference components on the allowable subspace at each receiver. Despite being a promising technique for the interference mitigation in the large wireless network, interference alignment is far from being practical. A key challenge in implementing existing interference alignment algorithms is that each of transmitters and receivers requires the global and perfect channel state information (CSI) of all interference channels, resulting in the overwhelming overhead.
This thesis proposes the efficient feedback design for implementing interference alignment in K-user MIMO interference channels, which provides a dramatic reduction of feedback overhead compared with the conventional feedback
approach. Based on the proposed feedback schemes, we consider the finite-rate feedback between transmitters and receivers and analyze the resultant residual interference due to the imperfect CSI at each transmitter. Furthermore, the adaptive feedback-bits allocation algorithm that minimizes a sum of residual interference is proposed under the constraint of total feedback bits.
Secondly, the clustered interference alignment is considered as the alternative method of interference mitigation with low-level feedback overhead in wireless ad-hoc network. To reduce the strong interference from neighboring transmitters, we group the transmitter-receiver pairs located close together and perform the interference alignment under the local CSI exchange between neighboring transmitters.
Since the clustered interference alignment inevitably provides the intercluster interference, the efficient inter-cluster interference mitigation scheme that exploits the characteristics of cluster geometry is proposed.
Finally, interference alignment in the cellular network is studied under the limited feedback and backhaul constraint. The upper-bound of residual interference due to the misaligned interferers is analyzed using random vector quantization and the optimal reference vector to be aligned with other cell interference is derived, which improves the system performance compared with the conventional interference alignment in the cellular network.