Technical Field

The exemplary and non-limiting embodiments of this invention relate generally to wireless communications and more specifically to precoder vector selection (e.g., in LTE systems).

Background Art

The following abbreviations that may be found in the specification and/or the drawing figures are defined as follows:

3 GPP third generation partnership project

CQI channel quality indicator

CRS common (cell-specific) reference signal

CSI channel state information

CSI-RS channel state information reference signal

DL downlink

DMRS demodulation reference signal (user specific)

E-UTRA evolved universal terrestrial radio access

eNB, eNodeB evolved node B /base station in an E-UTRAN system

ePDCCH enhanced physical downlink control channel

eREG eResource element group

E-UTRAN evolved UTRAN (LTE)

LTE long term evolution

LTE-A long term evolution advanced

NCT new carrier type

PDSCH physical downlink shared channel

PUCCH physical uplink control channel

PUSCH physical uplink shared channel

PMI precoding matrix index

PRB physical resource block

RAN radio access network

SCH synchronization channel

SFBC space-frequency block code

SNR signal-to-noise ratio SINK. signal to interference plus noise ratio

SRS sounding reference signal

UE user equipment

UL uplink

ULA uniform linear array

UTRAN universal terrestrial radio access network

WG working group

In LTE 3GPP Release- 1 1, ePDCCH is supported using UE-specific demodulation reference signals (DMRS) instead of common reference signals (CRS), e.g., see 3GPP-RAN WG 1 Meeting #68, Rl- 120109, Link level evaluation on E-PDCCH transmission schemes, CATT, February 6-10, 2012, Germany, and 3GPP-RAN WG1 Meeting #68, Rl- 120628, DM-RS Utilization of Diversity and Beamforming Schemes for ePDCCH, MediaTek Inc., February 6-10, 2012, Germany. In addition, in an upcoming 3GPP LTE Release 12, a new carrier type (NCT) will be introduced that will exclusively use DMRS for PDSCH demodulation. In both cases, it is assumed that the DMRS will be beamformed for each individual UE.

Summary

According to a first aspect of the invention, a method comprising: selecting, by a network element using a statistical predefined criterion, a plurality of precoding vectors with a probabilitydetermined for each selected precoding vector; and signaling by the network element to the user equipments using the selected precoding vectors based on the determined probability for each selected precoding vector.

According to a second aspect of the invention, an apparatus comprising:_a processing system comprising at least one processor and a memory storing a set of computer instructions, in which the processing system is arranged to cause the apparatus to:_selecting, using a statistical predefined criterion, a plurality of precoding vectors with a probability determined for each selected precoding vector; and signaling by the network element to the user equipments using the selected precoding vectors based on the determined probability for each selected precoding vector.

According to a third aspect of the invention, a computer program product comprising a computer readable medium bearing computer program code embodied herein for use with a computer, the computer program code comprising: code for selecting, by a network element using a statistical predefined criterion, a plurality of precoding vectors with a probability using signals received from user equipments, determined for each selected precoding vector; and code for signaling by the network element to the user equipments using the selected precoding vectors based on the determined probability for each selected precoding vector.

Brief Description of the Drawings:

For a better understanding of the nature and objects of embodiments of the invention, reference is made to the following detailed description taken in conjunction with the following drawings, in which:

Figure 1 is a diagram demonstrating a cell with a non-uniform user distribution (e.g., with hot-spots);

Figure 2 is an example of a preferred precoding vector distribution for 4 transceivers with ULA antennas, according to an exemplary embodiment of the invention;

Figure 3 is an example of a preferred precoding vector distribution for 4 transceivers with 2 pairs of cross-pole antennas, according to an exemplary embodiment of the invention;

Figure 4 is a flow chart demonstrating an exemplary embodiment of the invention performed by an eNB;

Figure 5 is another flow chart demonstrating an exemplary embodiment of the invention performed by an eNB; and

Figure 6 is a block diagram of wireless devices for practicing exemplary embodiments of the invention.

Detailed Description

To support beamforming of the DMRS and associated control and data channels, two approaches may be considered. First, the network can obtain knowledge of the propagation channels either explicitly (e.g., using a sounding reference signal SRS) or through an appropriate feedback (e.g., using a covariance matrix). Alternately, a UE can directly feedback a preferred precoding matrix indicator (PMT) to be used by the network. In certain situations, however, this is not possible (e.g., during initial attachment of the UE to a cell). Furthermore, the channel estimates or feedback may be of poor quality in some cases and therefore unusable for beamforming. Therefore, a fallback mode is desirable. In previous 3GPP LTE releases, SFBC was used as the fallback mode, but this is not possible to support the beamforming of the DMRS due to the absence of the CRS.

One potential fallback mode may involve precoder cycling (i.e., random beamforming). In this method, several predefined precoding vectors could be available (for example, 4 precoding vectors) and the network may cycle through the available precoding vectors (e.g., 0, 1 , 2, 3, 0, 1, 2, 3,...). For instance, each precoding vector may be applied to one physical resource block or one eREG (eResource Element Group) depending on how the transmission is structured. This method then may be transparent to the UE and under exclusive implementation control of the eNB. It will not require any change to 3GPP specifications. However, simply cycling through all available precoding vectors may not be an optimal solution. This is because certain precoding vectors may be preferred over others. That can be due to, for example, cell deployments, user distribution, directional coverage, antenna configuration and other factors.

Figure 1 shows an example of different concentrations of UEs in a macro-cell 10 with eNB 11 where UEs in geographical areas 12 and 14 have high concentration, so that the UEs in the area 12 may select one preferred precoding vector, the UEs in the area 14 may select another preferred precoding vector, and the preferred precoding vectors for the UEs in the area 12 and 14 may be different than for other UEs, in the cell 10.

A new method, apparatus, and software related product (e.g., a computer readable memory) are presented for precoding vector selection (e.g., in LTE wireless systems). Instead of cycling through all available precoding vectors, a network element such as eNB may select precoding vectors based on a prioritized distribution of precoding vectors within the cell using initialization with a default codebook. This ensures that preferred precoding vectors are selected more often than other precoding vectors. The method is of low-complexity and intended to ensure a reliable performance of devices in LTE networks. The embodiments described herein are implementation-specific and are done exclusively at the network element (access point) such as eNB. According to embodiments of the invention, the network element may select, using a statistical predefined criterion, a plurality of precoding vectors with a determined probability based on information received from UEs, where the probability is determined for each selected precoding vector. Then the network element may send data/control signals to the UEs using selected plurality of the precoding vectors based on the determined probability for each selected precoding vector.

Figures 2 and 3 illustrate a preferred precoding vector distribution in a cell with different antenna configurations: for 4 transceivers with ULA antennas in Figure 2 and for 4 transceivers with two pairs of cross-pole antennas in Figure 3. The precoding vectors are selected from the 3GPP codebook for four transmit antennas.

From Figures 2 and 3, it can be ascertained that only a few vectors are preferred. For example, from Figure 2 it is seen that precoding vectors with indices 0, 1 , 4, 5, 7, 8 are preferred. Furthermore, a precoding vector with the index 1 is by far the most preferred precoding vector by users in the cell. Conversely, some precoding vectors are rarely selected. Similarly, in Figure 3 the precoding vector with the index 1 is also the most preferred precoding vector by users in the cell. Therefore, a simple cycling through all available precoding vectors is not an optimal way to select precoding vectors and can result in suboptimal performance, so that embodiments described herein can offer a better solution.

According to a first embodiment, the preferred precoding vectors may be determined/selected based on channel information knowledge. The network element (eNB) can start with a default codebook and collect statistics regarding the preferred precoding vectors within the codebook (the precoding vectors corresponds to codewords from the default codebook) using, for example, sounding reference signals (SRS) from the UEs in the cell. This can be done at the eNB once channel information is known (e.g., by selecting the precoding vector that will maximize SNR, SINR or throughput, etc.). In other words, the eNB may determine/rank, using the collected statistics from the UEs, a first preferred precoding vector with ranklp, a second preferred precoding vector with rank2p (i.e., next preferred if the first in not available), a third preferred precoding vector with rank3p, etc. for each SRS from the UE (see Table 1 for details as discussed below). This ranking may be performed by the eNB using a predefined quality parameter, e.g., by computing SNR, SINR or throughput for all available precoding vectors in the SRS.

Further, the eNB may determine/calculate a score of a given precoding vector based on collected statistics including, for example, how many times this precoding vector was the first, second, third, etc. vector. This score can be calculated using the following equation:

( 1), where s, is a score for a vector /, w¾ is a weight for k-t preferred vector and tiu is a preferred vector count for the vector . The Equation 1 may be further refined as a function time t. For example, if a weighted histogram is used, more weight may be assigned for recent subframes and less weight may be assigned for subframes in the past. Then the Equation 1 may be further modified, e.g., using a time scaling factor a, as follows: si =∑ _k e ^at w _k n _{k i} (2).

Moreover, the eNB may determine/calculate a probability p, of selecting a given precoding vector based on the determined vector scores as follows: Then based on the calculated probability, the eNB may use any given precoding vector i with the probability p; for DL transmission (data and/or control) to the UEs. Note that according a further embodiment, the eNB may remove some precoding vectors from the codebook if their selection probabilities are zeros or below a predefined threshold. In LTE, these precoding vectors can be applied to DL channels such as PDSCH, ePDCCH, and SCH. The precoding vectors may be applied to the entire transmission or to a part of the transmission. For example, if a UE is given an assignment of 10 PRBs in the PDSCH, for each PRB, a different precoding vector can be applied. Thus for each PRB, the network can pick a precoding vector to be applied based on the probability. In another example, for the ePDCCH, a different precoding vector can be applied for each eREG.

Thus according to the first embodiment, as described herein, the network element (eNB) may receive UL reference signals (e.g., SRSs) from the UEs and rank precoding vectors for each received uplink reference signal using a predefined quality parameter such as SINR, calculate a score for each precoding vector based on the ranking of each precoding vector for each received uplink reference signal, and determine the probability of each precoding vector based on the calculated scores for all precoding vectors.

An example illustrating the first embodiment is shown in Table 1 below where 10 UEs are in the system/cell periodically send sounding reference signals (SRS) to the eNB. The eNB calculates preferred precoding vectors after every SRS transmission. After 1 1 SRS transmissions, the eNB calculates the distribution as shown below. In Table 1, the eNB calculates the preferred precoding vectors as Vj, v _j where v; is a first preferred precoding vector (with assigned probability of 0.8), and Vj is a second preferred precoding vector (with assigned probability of 0.2) for each SRS transmission.

Table 1. Statistics collected by the eNB according to the first embodiment.

v ₂ , V3, v ₂ , V3, V6, v ₂ , V3, 3, V3, v ₂ , v ₂ ,

3

Vl Vl V l V ₂ v ₂ V3 v ₂ v ₂ v ₂ V3 v ₃ v ₂ , l , v ₂ , Vl , v ₂ , V6, V3, v ₂ , v , v ₂ , v ₂ ,

4

l v ₂ V l V2 V3 ₂ v ₂ v ₂ V3 V3 V3 v ₂ , v ₆ , v ₂ . V6, V2, V l , v ₂ , v ₂ , V2, v ₂ , v ₂ ,

5

Vl Vl V| v ₂ V3 v ₂ V3 v ₂ V 3 V3

V2, V2, v ₂ , V7, v ₂ , v ₂ , v ₂ , v ₂ , V2, v ₂ , v ₂ ,

6

Vl Vl Vl V ₂ v ₃ V3 V3 v ₂ V3 V3 v ₃

V2, v ₂ . V6, V3, v ₂ , v ₂ , V3, Vl , v ₂ , V l , v ₂ ,

7

Vl Vl Vl 2 3 V3 v ₂ v ₂ v ₃ V2 V3 v ₂ , ^v 2, V3, V3, v ₂ , v ₂ , V3, V3, V3, v ₃ , v ₂ ,

8

l Vl V l V ₂ V3 V3 v ₂ v ₂ v ₂ v ₂ V

V3, V6, V3, V3, V l , V3, v ₃ . v ₂ , V3, l , v ₂ ,

9

Vl Vl Vl V ₂ v ₂ V? v ₂ V3 v ₂ v ₂ V3

V3, V5 , V3, V7, V3, V3, v ₂ , ^v 2, V3, V3, v ₂ ,

10

V l Vl V l Vl V2 V ₂ V3 V3 v ₂ v ₂ V3

The calculation may be performed using Equations 1 and 3 as follows.

Assuming wo=0.8 and W]=0.2 and using Equation 1 , it can be found that so=0, si=0.8* 10+0.2*29,

s ₇ =0.8*2.

Thus, s ₀ =0, s,=13.8, s ₂ =47.8, s ₃ =32.4, s ₄ =0, s ₅ =0.8, s ₆ =5.6, s ₇ =1.6.

Then corresponding probabilities calculated using Equation 3 are: po=0.000, pi=0.135, p ₂ =0.468, p ₃ =0.318, p ₄ =0.000, p ₅ =0.008, p ₆ =0.055, p ₇ =0.016. Thus the corresponding probabilities of using precoding vectors are the following: for vo it is 0%, for vi it is 13.5%, for v ₂ it is 46.8%, for v ₃ it is 31.8%, for v ₄ it is 0%, for v ₅ it is 0.8%, for v ₆ it is 5.5%, for v ₇ it is 1.6%. Then if we use 1% as a threshold probability for using the precoding vectors, the eNB may remove precoding vectors vo, v ₄ , V5 from the codebook.

According to a second embodiment, the preferred precoding vectors may be determined/selected based on a PMI feedback from the UEs. The network element (eNB) can start with a default codebook and collect statistics regarding the preferred PMI reported by the UEs. Note that in one option only statistics about the first preferred precoding vector can be obtained from the PMI feedback signal from the UE, but in general it could be more than one preferred precoding vectors as further discussed herein.

Next, the eNB may determine/calculate a score of a given precoding vector based on collected statistics including, for example, how many times a particular vector was in the reported feedback (PMIs) by the UEs. This score for the statistics collected about the first preferred precoding vector can be calculated using the following equation:

Si -∑k ⁿ k,i (4), where s, is a score for a vector and /¾,, ^■ is a precoding vector count reported by the UEs for the vector . The Equation 4 may be further refined as a function time t. For example, if a weighted histogram is used, more weight may be assigned for recent subframes and less weight may be assigned for subframes in the past. Then the Equation 4 may be further modified, e.g., using a time scaling factor a, as follows:

Si =∑ _k e ^at n _kil (5).

Moreover, the eNB may determine/calculate a probability pj of selecting a given precoding vector based on the determined vector scores using Equation 3 as in the first embodiment.

Then, based on the calculated probability, the eNB may use any given precoding vector i with the probability pi for DL transmission (data and/or control) to the UEs. Note the eNB may remove some precoding vectors from the codebook if their selection probabilities are zeros or below a threshold as in the first embodiment.

Thus according to the second embodiment, as described herein, the network element

(eNB) may receive UL feedback signals (e.g., PMI signals) from the UEs, each feedback signal comprising information on one or more preferred precoding vectors, calculate a score for each precoding vector comprised in a default codebook using the received information, and determine the probability of each precoding vector based on the calculated scores for all precoding vectors.

An example illustrating the second embodiment is shown in Table 2 below where 10 UEs are in the system/cell periodically send PMI signals (e.g., as a part of a CSI feedback report) to the eNB and using only first preferred precoding vector in the PMI report by the UEs. After 1 1 PMI transmissions, the eNB calculates the distribution as shown below. Table 2. Statistics collected by the eNB according to the second embodiment.

The calculation may be performed using Equations 4 and 3 as follows.

Using Equation 1, it can be found that so=0, Si=10, s ₂ =60,

Then corresponding probabilities calculated using Equation 3 are: po=0.000, p ₂ =0.545, p ₃ =0.273, p ₄ =0.000, p ₅ =0.009, p ₆ =0.064, p ₇ =0.018. Thus the corresponding probabilities of using the precoding vectors are the following: for vO it is 0%, for vi it is 9.1%, for v _: it is 54.5%, for v ₃ it is 27.3%, for v ₄ it is 0%, for v ₅ it is 0.9%, for v ₆ it is 6.4%, for v ₇ it is 1.8%. Then if we use 1% as a threshold probability for using precoding vectors, the eNB may remove precoding vectors vo, v ₄ , v, from the codebook.

As it is noted herein, in general more than one preferred precoding vectors may be obtained from the PMI feedback signal from one UE. For example, if the UE in the CSI feedback may report two or more corresponding preferred precoding vectors, the eNB may use Equation 1 with assigned weights w¾ as explained in the first embodiment and Table 1.

Alternatively the w information for each preferred precoding vector can be asserted based on CQI information for each preferred precoding vector which can be contained in the CSI feedback report.

Furthermore, in the 3GPP LTE Releases 8-11 precoders have the nested structure where a high rank precoder is composed from low rank precoders.

In case the PMI feedback from a UE is for a desired precoder with rank >1 , then the rank=l precoders forming the desired precoder with the rank >1 are considered desirable for statistics accumulation.

For example, the eNB may have 4 antenna ports so that the UE can feedback a rank 3

PMI, which means that the preferred precoder is a 4x 3 matrix [vi v ₂ v ₃ ], . In LTE, each column vector of a precoder matrix at rank > 1 is actually a rank 1 precoder in the codebook. Then the indices for precoders vi, v ₂ and V3 may be included in the statistices (e.g., see Tabl;e 2). In this example, vi, v ₂ and V3 are 3 rank 1 (4 x 1 vectors ) precoders in the codebook. As the UE prefers a rank 3 precoder (a matrix), and the random precoders used for control channel are rank 1 precoders, the question may be "which rank 1 precoder(s) will the UE prefer?". In this case, the answer may be that each vector may be assigned an equal weight with a score of 1/3.

In the future (e.g., 3GPP LTE Release 12/13), if codebook design includes non-nested codewords, then subspace-fitting can be used to approximate the desired precoder. For example, the desired high rank precoder is [vi, v ₂ , v ₃ ], but vi, v ₂ , V3 themselves are not included in the codebook. Then three rankl vectors from the codebook wi, w ₂ , w ₃ may be searched so the subspace spanned by [vi, v , v ₃ ] can be found to be close to the subspace spanned by [wi , w ₂ , W3] as further explained below.

As stated above, there is no guarantee that in the future the codebook still has the property so that a column vector of a high rank precoder is actually a legitimate codeword in the (default) codebook. For example, even though the matrix V= [vi, v ₂ , v ₃ ] is in the codebook CB= { wi , w ₂ , w ₃ , W4, ... , A, B, ... , V, ... } which may include rank 1 precoders, rank 2 precoders, rank 3 precoders, etc., yet vi, v ₂ and v ₃ are not in the codebook CB. Then we can find some rank 1 precoders in the codebook CB such as wi, w ₂ and W3 which may be close to forming the matrix V. For example, one can examine the projection of w, into V : || w"V \\ , where || · || is a Frobenius norm, and » stands for a conjugate transpose; a large value of this projection may suggest that w _t is really covered by V . But our real goal is not to find a rank precoder well covered by V, rather we are interested in finding a couple of rank 1 precoders to cover V. In linear algebra, the concept of distance between subspaces has been developed for a long time. This concept can be used to measure how well a chosen set [w, w ₂ w ₃ ] covers V by considering the subspace distance between the matrix [w _x w, w _} ] and V. Once we have found a good fit (in this case {w _] , w ₂ , w ₃ } ), then we can effectively take [w, w ₂ ¼>,] as the preferred precoder by the UE, then we can give , , w ₂ , \v _} a weight of 1 /3 as explained above.

In another scenario applied in the second embodiment described herein, the eNB may use different downtilts to the CSI-RS resources without the knowledge of the UE so that the UE can use these CSI-RS resources for the CSI feedback which includes the PMI feedback. This further scenario is disclosed in a co-owned PCT patent application "CSI Feedback With Elevation Beamforming", International Application No. PCT US2012/061082 filed on October 19, 2012.

It is noted that a further scenario can be used in the second embodiment described herein, when the eNB may apply some precoding vectors to the CSI-RS resources without the knowledge of the UE so that the UE can use these CSI-RS resources for the CSI feedback which includes the PMI feedback. This further scenario is disclosed in a co-owned PCT patent application "Codebook Construction Using CSI Feedback For Evolving Deployment

Scenarios", International Application No. PCT/US2012/054409, filed on September 10, 2012.

Furthermore, the preferred precoder may be completely described by the UE feedback and transmission choices at the eNB side. This approach for non-nested codebook can be applied using a covariance matrix based design as explained below. An example of a general case of the covariance matrix based design where the eNB can apply some precoding vectors to the CSI-RS resources is provided below.

For instance, all UEs may be required to send CSI reports to an eNB at subframes 1, 6, 11, 16 and 21, so that precording vectors, e.g., corresponding to unitary matrices Mi, M ₂ , M ₃ , ...., may be applied to the CSI-RS at subframes 1, 6, 1 1, 16 and 21. Suppose that the eNB applies a unitary matrix M _\ at subframe 1 which may be given by:

-0.5705 - 0.3613 - 0.1621 0.7196

- 0.3061 0.1444 0.9401 0.0416

M,

-0.6310 - 0.3268 -0.1247 -0.6924

-0.4274 0.8613 - 0.2729 0.0321

Further, the unitary matrix applied at subframe 6 may be given by:

^" - 0.6092 -0.6306 0.1504 0.4566

- 0.5193 0.4978 0.6582 - 0.2222

-0.5042 - 0.1248 -0.5297 - 0.6705

- 0.3240 0.5822 - 0.5134 0.5408

, etc.

Then for a UE1, the eNB may find that the received CQI corresponding to PMI=

[ 1 - 1 - 1 - 1]' at subframe 1 at subbands k=l, 2, 3, 4, 5 and 6 is the highest among all CSI reports from the UE1 , so that the best precoding vector for the UE l is U _{X k} = M _x [ \ - 1 - 1 - 1 ]' at subbands k = 1, 2, 3, 4, 5 and 6.

In general, we can use U _{l k} = [U _x - - - U _k ] for the preferred precoding vector for UEi

' ' i.k

at subband k , which includes both the PMI feedback (PMI= [ 1 - 1 - 1 - 1 ] ^r ) and the precoding vector (e.g., M for the subframe 1 ) applied at the eNB side during the UE's CSI measurement.

Here r _{i k} is a rank indication feedback from the UEi for the subband k .

As a result of this approach the effective size of the codebook may be enlarged, e.g., proportionally to a number of different matrices applied by the eNB to the CSI-RS. It may happen that different UEs have different preferences, and no precoding vector (or a set of precoding vectors) receives an overwhelming score. Then a scheme is needed for the eNB to derive a relatively small set of precoding vectors. Now the problem can be formulated as finding a subspace or a set of precoding vectors (which may not necessarily be the basis vectors for a subspace, so thatthe precoding vectors may not be necessarily orthogonal to each other) to detemine the preferred precoding vectors for all the UEs at all the subbands.

In other words, the eNB needs to select a set of precoders so that, e.g., the broadcast information can be conveyed to all the UEs using that set of precoders. One possibility is to form a subspace approximation to the space spanned by the column vectors of U _{i k} for all the

UEs and the subbands with feedback. The following scheme may be used:

a) Form a coveriance matrix R - _k U" _k , where » is a conjugate transpose

i.k

operator. The covariance matrix can provide indication on average what are the preferred precoding vectors.

b) Perform Eigen decomposition on R : and r _x , r ₂ , r ₃ , r ₄ (assume there are 4 Tx antenna ports from the eNB) may be the found eigenvectors, and the corresponding eigenvalues are d _x ≥d ₂ ≥d _} ≥d ₄ . Procedures for eigen-decomposition can be found, for example, on pp.391 -469, ^{" '} Matrix Computations", The Johns Hopkins University Press, (3rd Edition), by Gene H. Golub and Charles F. van Van Loan, 1996.

c) r corresponds to the single precoding vector which can serve best all the UEs on average, yet some UEs will suffer if r , which is the eigenvector corresponding to the largest eigenvalue d _x , is not well aligned with their preferred precoding vectors. r ₄ , which is the eigenvector corresponding to the smallest eigenvalue d ₄ , is the single precoding vector which is least favored by all UEs on average, yet a minority UEs' preferred precoding vectors may be well aligned with r ₄ which may serve them quite well. If we use {r _x } as a set of precoders, then some UEs such as UEj 1 may be served well if the correlation between r and U _{;l k} is high; and some UEs such as UEj2 may be served poorly if the correlation between r _x and U _{/2 k} is low. Then we can expand the set of precoders from {r _x } to {r _x , r ₂ } with the hope that the worst off UE is helped in this process. And this process can continue to expand the set of precoders to {r,,r ₂ ,r ₃ ,r ₄ } . The score for each eigenvector is given by d _i , i.e. s _; = d _t . And the probability of transmitting eigenvector η is given by

_P , ί

of selected eigenvectors) (&) d) The transmit power of each precoder can be also adjusted to achieve a trade-off between UEs. For example, if ,r ₂ ,r _} ,r ₄ are all included in the selected precoding vectors, the transmit power for r _x ,r ₂ ,r ₃ ,r ₄ can be adjusted according to two conflicting goals of serving

UEs in general well and serving the worst off UEs well, e.g., r ₄ may be given more power than e) The probability of transmission of each precoding vector can be also adjusted to reconcile between the two conflicting goals of serving UEs in general well and serving the worst off UEs well. For example, by the goal of serving UEs in general, if ,r ₂ ,r ₃ ,r ₄ are all included in the selected precoding vectors, e.g., r ₄ may be transmitted as frequently as prescribed (see Equation 6) by P ₄ = -j 4 - ¹ — (e.g., P4=10%), and r _x is transmitted as frequently as prescribed j

(see Equation 6) by P _x = ^l (e.g., Pl=35%). However, considering the second goal of

serving the worst off UEs well, r ₄ may be transmitted more frequently than P ₄ = ₄ ⁴ (say -1 r ₄ is transmited at 15%) and r _{ may be transmitted less frequently than P _t - ₄ ¹ (say r, is

y-1

transmitted at 30%). f) We can also modify the covariance matrix by giving a weight to each U _{i k} as in the first embodiment: R = fV _{i li} U _{l k} U" _k , and follow the procedure given above.

It is noted that the above scheme with elements a) through f) may be applied to a simplified case of the CSI (PMI) feedback from the UEs without applying by the eNB precoding vectors to the CSI-RS, i.e., Mi, Mi, M3, .. ..being identity matrices in the above algorithm.

According to a third embodiment, the network element (eNB) can start with a default codebook and collect statistics regarding channel estimates, for example, using sounding reference signals (SRS) and/or preferred PMI feedback from the UEs (like in the first and/or second embodiments described herein). Then the eNB can construct a new codebook based on the collected channel information such as SRSs and/or appropriate feedback (e.g., PMI). This new codebook may be significantly smaller than the default codebook. Note that the new codebook should not be the same as 3GPP LTE Release-8 codebook. Then the eNB can use a scoring and determining selection probabilities (see Equations 1-5) for each precoding vector in the new codebook as described herein in reference to the first and second embodiments. In other words, the first or second embodiments may be applied to the precoding vectors in the new codebook.

Figure 4 shows an exemplary flow chart demonstrating implementation of embodiments of the invention by a network element (e.g., eNB). It is noted that the order of steps shown in Figure 4 is not absolutely required, so in principle, the various steps may be performed out of the illustrated order. Also certain steps may be skipped, different steps may be added or substituted, or selected steps or groups of steps may be performed in a separate application.

In a method according to the exemplary embodiment shown in Figure 4, in a first step 40, the network element (eNB) receives UL reference signals from UEs (e.g., in a cell served by the eNB) and ranks precoding vectors for each received uplink reference signal using a predefined quality parameter (e.g., SINR), the precoding vectors corresponds to codewords from a default codebook. In a next step 42, the eNB may generate (optionally) a reduced codebook by removing a number of codewords from the default codebook based on the ranking of the precoded vectors by selecting preferred one or more precoding vectors.

In a next step 44, the eNBs calculates a score for each precoding vector comprised in the default or reduced codebook based on the ranking of precoding vectors for each received uplink reference signal. In a next step 46, the eNB determines the probability of each precoding vector in the default or reduced codebook based on the calculated scores for all precoding vectors in the default or reduced codebook.

In a next step 47, the eNB selects a plurality of precoding vectors with the determined probability (e.g., some precoding vectors with zero probability or a low probability below a predefined threshold may be excluded). In a next step 48, the eNB sends data to the UEs using selected precoding vectors based on the determined probability for each selected precoding vector.

It is further noted that according to a further embodiment, steps 40-48 can be repeated, e.g., every predefined time period.

Figure 5 shows another exemplary flow chart demonstrating implementation of embodiments of the invention by the network element such as eNB. It is noted that the order of steps shown in Figure 5 is not absolutely required, so in principle, the various steps may be performed out of the illustrated order. Also certain steps may be skipped, different steps may be added or substituted, or selected steps or groups of steps may be performed in a separate application.

In a method according to the exemplary embodiment shown in Figure 5, in a first step 50, the network element (eNB) receives feedback signals from UEs, each feedback signal comprising information on one or more preferred precoding vectors. In a next step 52, the eNB may optionally generate a reduced codebook by removing a number of codewords from a default codebook based on the feedback signals received from the UEs. In a next step 54, the eNB calculates a score for each precoding vector comprised in a default or reduced codebook using the received information. In a next step 56, the eNB determines the probability of each precoding vector in the default or reduced codebook based on the calculated scores for all precoding vectors in the default or reduced codebook.

In a next step 57, the eNB selects a plurality of precoding vectors with the determined probability (e.g., some precoding vectors with zero probability or a low probability below a predefined threshold may be excluded). In a next step 58, the eNB sends data to the UEs using selected precoding vectors based on the determined probability for each selected precoding vector. It is further noted that according to a further embodiment, steps 50-58 can be repeated, e.g., every predefined time period.

Figure 6 shows an example of a block diagram demonstrating wireless devices (e.g., in LTE) including a network element (e.g., eNB) 80 comprised in a network 100, and a UE 82 communicating with the eNB 80, according to an embodiment of the invention. Figure 6 is a simplified block diagram of various electronic devices that are suitable for practicing the exemplary embodiments of this invention, and a specific manner in which components of an electronic device are configured to cause that electronic device to operate. The UE 82 may be a mobile phone, a camera mobile phone, a wireless video phone, a portable device or a wireless computer, etc.

The eNB 80 may comprise, e.g., at least one transmitter 80a, at least one receiver 80b, at least one processor 80c at least one memory 80d and a precoder vector selection application module 80e. The transmitter 80a and the receiver 80b may be configured to provide a wireless communication with the UE 82 (and others not shown in Figure 6), e.g., through a corresponding link 81, according to the embodiments of the invention. The transmitter 80a and the receiver 80b may be generally means for transmitting/receiving and may be implemented as a transceiver, or a structural equivalence thereof. It is further noted that the same requirements and considerations are applied to transmitter and receiver of the UE 82.

Various embodiments of the at least one memory 80d (e.g., computer readable memory) may include any data storage technology type which is suitable to the local technical environment, including but not limited to semiconductor based memory devices, magnetic memory devices and systems, optical memory devices and systems, fixed memory, removable memory, disc memory, flash memory, DRAM, SRAM, EEPROM and the like. Various embodiments of the processor 80c include but are not limited to general purpose computers, special purpose computers, microprocessors, digital signal processors (DSPs) and multi-core processors. Similar embodiments are applicable to memories and processors in other wireless devices such as UE 82 shown in Figure 6.

The precoder vector selection application module 80e may provide various instructions for performing steps 40-48 shown in Figure 4 and steps 50-58 shown in Figure 5. The module 80e may be implemented as an application computer program stored in the memory 80d, but in general it may be implemented as software, firmware and/or hardware module or a combination thereof. In particular, in the case of software or firmware, one embodiment may be implemented using a software related product such as a computer readable memory (e.g., non-transitory computer readable memory), computer readable medium or a computer readable storage structure comprising computer readable instructions (e.g., program instructions) using a computer program code (i.e., the software or firmware) thereon to be executed by a computer processor. Furthermore, the module 80e may be implemented as a separate block or may be combined with any other module/block of the device 80, or it may be split into several blocks according to their functionality.

The UE 82 may have similar components as the eNB 80, as shown in Figure 6, so that the above discussion about components of the eNB 80 is fully applicable to the components of the UE 82.

It is noted that various non-limiting embodiments described herein may be used separately, combined or selectively combined for specific applications.

Further, some of the various features of the above non-limiting embodiments may be used to advantage without the corresponding use of other described features. The foregoing description should therefore be considered as merely illustrative of the principles, teachings and exemplary embodiments of this invention, and not in limitation thereof.

It is to be understood that the above-described arrangements are only illustrative of the application of the principles of the present invention. Numerous modifications and alternative arrangements may be devised by those skilled in the art without departing from the scope of the invention, and the appended claims are intended to cover such modifications and

arrangements.