Dynamic Coordinated multipoint transmission schemes

In LTE-Advanced, Coordinated MultiPoint (CoMP) transmission is one of the techniques proposed to mitigate intercell interference, especially for cell-edge users. CoMP techniques are divided into coordinated beamforming/scheduling and joint processing. This thesis focuses on joint processing, where the user receives its data from various base stations, improving the signal strength and canceling interference. Coherent joint processing imposes perfect channel knowledge and perfect synchronization between base stations, but yields substantial theoretical gains. In the previous work, three joint processing approaches were studied in a static cluster of base stations for a flat fading Rayleigh channel. In the Centralized Joint Processing approach, global channel state information was available at the transmitter side, and the base stations within the cluster jointly performed the power allocation and the design of the beamformer. This puts tremendous requirements on backhauling. The Partial Joint Processing approach formed a set of base stations within a predefined threshold for transmission, reducing the requirements in backhauling and feedback from users. Finally, in the Distributed Joint Processing scheme, the power allocation and beamformers were locally calculated for every base station. In this thesis, the performance of these algorithms is evaluated in a multipath environment using the WINNER II channel model. The worst case scenario in terms of interference is considered where all the users are allocated in all the resource blocks. Hence, the joint processing schemes are applied in the frequency domain in every resource block. In particular, the performance of the Partial Joint Processing algorithm is improved with frequency adaptive thresholding compared to non-adaptive frequency thresholding. The threshold values for the Partial Joint Processing algorithm depend on the WINNER II channel model. The relative average number of active links with frequency adaptive thresholding is lesser compared to non-adaptive thresholding. Fewer active links translate to sparse channel matrices available at the central unit and poses problems to design the zero-forcing beamformer. A partial zero-forcing method performs better under these conditions. For adaptive thresholds greater than 20dB, there is multiuser interference and the performance of the Partial Joint Processing scheme degrades when moving towards the cell-edge. In addition, the channel correlation matrix suffers from rank deficiency. This is more prominent near the base station. Based on this, an algorithm is developed which defines cooperation areas over the cluster as to when the Partial Joint Processing scheme can be applied or fallback on the Centralized or Distributed Joint Processing.