An Overbooking Strategy to Maximize the Attainable Load What is described in this subsection is an optional part of the JT CoMP concept. It is at present under study as a way of maximizing the attainable load and thus the attainable spectral efficiency and throughput at high loads. Already mentioned above was that per cell sets of e.g. 3 UEs in a 4-antenna cell should be found. This we call an “optimistic overbooking” strategy. With four antenna elements per cell, a maximum of 4 data streams per cell would be possible, but it is well known at most 3 users should in practice be served. In single-cell designs, it is hard to serve 3 users per PRB in 4- antenna cells successfully. The “optimistic overbooking strategy” strives to attain this high number of users per cell in a JT CoMP setting. For uniform linear arrays (ULA) with λ/2 antenna spacing or similarly with λ-spacing, relatively broad wideband beams will be generated, that have a strong mutual overlap. In case more than two UEs should be served simultaneously, there will then be a badly conditioned channel matrix. Therefore, when trying to serve 3 UEs per cell simultaneously one has to be optimistic, and the hope is that the CA-wide JT CoMP precoding will result in a cooperation area-wide channel matrix with acceptable eigenvalue spread. Relative to what is normally possible by single cell MU-MIMO transmission, the here proposed system is thus ‘overbooked’ with 3 UEs per cell, all being in CoMP mode. Only in the end some fine tuning is done where needed. This aggressive cell loading strategy guaratees that all potential COMP gains obtainable for a given number of spatial degrees of freedom are exploited and that the penetration rate of CoMP UEs becomes near to 100%.17 This overbooking strategy at present works only in combination with an extensive per-cell optimization, including an exhaustive search for all possible sets of 3 user groups (e.g. 120 potential groups in case of 10 UEs), adapting the UE filters and finding the best allocation of UE sets to PRBs. For the last task, a per-set instead of per-UE Score-based scheduler has been found to be useful as it will lead to a good degree of fairness. An important property that limits the performance of joint linear precoders in too highly loaded cells or cooperation areas is the power normalization loss problem: With a fixed number of transmit antennas and an increasing number of UEs per PRB, linear zero-forcing precoders, which perform channel inversion, will tend to get larger matrix el...
