Robust Interference Management for Ultra-Dense Heterogeneous Networks

• Main Authors: Tai, Shao-Heng, 戴紹恆
• Other Authors: Fung, Carrson C.
• Format: Others
• Language: en_US
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/tktb9b

碩士 === 國立交通大學 === 電子工程學系 電子研究所 === 104 === Performance of precoder based spatial intercell interference management in heteroge-neous networks can be hindered by a lack of accurate channel state information. Many works have shown designing precoder considering CSI mismatch models can bring perfor-mance improvement when CSI error is present. Sparsity enhanced mismatch model (SEMM) precoder has been proposed recently to exploit the dual parameter SEMM which in turn ex-ploits the inherent sparse characteristics of MIMO interference channels and was shown to have better performance compared to the conventional norm ball mismatch model (NBMM) precoder. However, when communicating and interference channels are highly correlated, which will be defined in the sequel, performance of the SEMM precoder degrades and in some cases, it will underperform the NBMM precoder. A two-stage precoding scheme, which includes a second-stage precoder, is proposed herein to circumvent this problem. This approach attempts to “orthogonalize” the communicating (H) and interfering channels (Gk, with k as the victim index) and is shown by simulations to outperform the SEMM pre-coder in high channel correlation scenarios. Even though the proposed precoding tech-nique does not orthogonalize the mismatch model nor truly orthogonalize the channels, for notation convenience, the proposed precoding scheme shall be referred ``orthogonalized'' SEMM precoding, or OSEMM precoding. An analysis will also be given to explain such performance gain brought forth by the OSEMM precoder. In addition, even though the proposed precoding scheme is a SU-MIMO precoding scheme, using the proposed channel correlation definition, the scheme can be used as a method for selecting the best user, i.e. user with the best channel, in a multiuser, multi-victim network, such that optimal perfor-mance can be attained for the targeted user.