Efficient Backhauling in Cooperative MultiPoint Cellular Networks

Sammanfattning: The efficient use of the spectrum in cellular systems has given rise to cell-edge user equipments (UEs) being prone to intercell interference. In this regard, coordinated multipoint (CoMP) transmission is a promising technique that aims to improve the UE data rates. In a centralized network architecture, the users need to feed back the channel state information (CSI) to its anchor base station (BS). The CSI is then forwarded to a central coordination node (CCN) for precoder design to jointly mitigate interference. However, feeding back the CSI consumes over-the-air uplink resources as well as backhaul resources. To alleviate this burden, the quantity of CSI being fed back is limited via relative thresholding. That is, the CSI feedback is limited to those BSs whose signal strength fall above a threshold relative to the strongest BS. Moreover, with limited CSI, efficient backhauling of the precoding weights is necessary, as the user data is routed based on the path taken by the precoding weights from the CCN to the corresponding BSs. The focus of this thesis is mainly on a physical (PHY) layer and a medium access control (MAC) layer approach for reducing the backhaul load in a CoMP system, with minimal penalty on the potential CoMP gains. Furthermore, broadcasting the CSI in a decentralized network architecture is considered in order to reduce backhaul latency. In the PHY layer approach, the precoder design is based on stochastic optimization such as particle swarm optimization (PSO). This method has no constraints on the scheduling of the UEs. The PSO based precoder design was also applied to field measurement data with CSI imperfections due to prediction errors and quantization errors. It was found to perform the best compared to other robust precoders developed in the EU FP7 ARTIST4G consortium. With the MAC layer approach, a simple zero forcing precoder is assumed, which focuses on how to schedule the UEs in such a way that they achieve the backhaul load reduction. Lastly, the decentralized network architecture is explored, where the UEs broadcast the CSI. The BSs coordinate by sharing minimal scheduling information, thereby achieving data rates comparable to the centralized network architecture. In this thesis, the backhauling is defined to be efficient when the total number of CSI coefficients aggregated at the CCN is equal to the total number of precoding weights for a given time-frequency resource, in a centralized architecture with the PHY layer approach. In the MAC layer approach, the total number of precoding weights is less than or equal to the total number of CSI coefficients. In the decentralized network architecture, the CCN does not exist. The BSs can coordinate over a less stringent backhaul, thereby reducing the backhaul load and latency.