Generating precoders for joint transmission from multiple transmission points to multiple user equipments in a downlink coordinated multipoint transmission/reception communications system

Abstract

There is provided generating precoders for joint transmission (JT) in a downlink coordinated multi-point transmission/reception (DL COMP) wireless communications system. The system includes a plurality of transmission points (TPs) operable to communicate with a plurality of user equipments (UEs). Each UE has one of the TPs as its serving TP. The method includes transmitting channel state information (CSI) from each UE to its serving TP, wherein the transmitted CSI includes precoder matrix indicators (PMI), and using the PMI to generate precoders for transmission of data from the plurality of TPs to the plurality of UEs.

Claims

1. A method for generating precoders for joint transmission (JT) in a downlink coordinated multi-point transmission/reception (DL CoMP) wireless communications system, the system including a plurality of transmission points (TPs) operable to communicate with a plurality of user equipments (UEs) wherein each UE has one of the TPs as its serving TP, and the method comprises: transmitting channel state information (CSI) from each UE to its serving TP, wherein the transmitted CSI includes precoder matrix indicators (PMI), and using the PMI to generate precoders for transmission of data from the plurality of TPs to the plurality of UEs, wherein using the PMI to generate precoders involves using the PMI to find a representative matrix (Ĥ.sub.in) representing the channel (H.sub.in) between an n-th TP and an i-th UE, wherein a fixed codebook (Ω.sub.RI) of representative matrices is generated from PMI codebook(s), the CSI transmitted from each UE to its serving TP includes a rank indicator (RI), and Ω.sub.RI is different for different RI, and if the RI for the i-th UE (RI.sub.i) is equal to the number of receive antennas of the UE (N.sub.RX) (i.e. if RI.sub.i=N.sub.RX) then Ω.sub.RI contains matrices Ĥ(m), m=1, . . . of size N.sub.RX×τ.sub.n, where τ.sub.n is the number of antennas at the n-th TP, i=1, . . . , N.sub.UE, n=1, . . . , N.sub.TP where N.sub.TP and N.sub.UE are the number of TPs and UEs, respectively.

2. The method as claimed in claim 1, wherein, if RI.sub.i is less than N.sub.RX (i.e. if RI.sub.i<N.sub.RX) then Ω.sub.RI contains vectors ĥ(m), m=1, . . . of size τ.sub.n×1.

3. The method as claimed in claim 2 wherein, for RI.sub.i=N.sub.RX, the representative matrix Ĥ.sub.in is found by:
Ĥ.sub.in=Ĥ(m*)εΩ.sub.RI,i=1, . . . ,N.sub.UE,n=1, . . . ,N.sub.TP with $m^{*}=\underset{m}{\arg \max }\mathrm{tr}\left\{{\left[\hat{H}\left(m\right)W_{in}\right]}^H\left[\hat{H}\left(m\right)W_{in}\right]\right\},\hat{H}\left(m\right)\in {\Omega }_{\mathrm{RI}}$ where N.sub.TP and N.sub.UE are the number of TPs and UEs, respectively, and W.sub.in (size τ.sub.n×RI.sub.i) is a precoder associated with a reported PMI used for precoding data to send from the n-th TP to the i-th UE according to 3GPP standard (TS 36.211).

4. The method as claimed in claim 3 wherein, for RI.sub.i<N.sub.RX, the representative matrix Ĥ.sub.in is found by: a) calculating correlation values as: $C_{\mathrm{in}}\left(m\right)=\mathrm{tr}\left\{{\left[{\hat{h}}^H\left(m\right)W_{\mathrm{in}}\right]}^H\left[{\hat{h}}^H\left(m\right)W_{\mathrm{in}}\right]\right\},m=1,\mathrm{.Math.}$ and b) sorting to find the N.sub.RX largest correlation values C.sub.in(m.sub.1)>C.sub.in(m.sub.2)> . . . >C.sub.in(m.sub.N.sub.RX) and the corresponding vectors ĥ(m.sub.1), ĥ(m.sub.2), . . . , ĥ(m.sub.N.sub.RX) to form the channel matrix
Ĥ.sub.in=[ĥ(m.sub.1),{circumflex over (h)}(m.sub.2), . . . ,{circumflex over (h)}(m.sub.N.sub.RX)].sup.H

5. The method as claimed in claim 4, wherein non-coherent precoding is used and the method further comprises using the representative matrix Ĥ.sub.in, a Lagrange multiplier ν.sub.n and a noise variance estimate σ.sub.i.sup.2 to compute the precoders (V.sub.in).

6. The method as claimed in claim 5, wherein precoders V.sub.in are computed using an iterative procedure.

7. The method as claimed in claim 6, wherein precoders V.sub.in are computed using the following iterative procedure where (m) denotes the m-th iteration: a) initialize a quantity G.sub.in(m=0)=J.sub.in, i=1, . . . , N.sub.UE, where J.sub.in is a RI.sub.i×τ.sub.n matrix with the (a,b)-th element being zero for a≠b and 1 for a=b; b) compute V.sub.in(m+1) using G.sub.in(m) and the Lagrange multiplier ν.sub.n for i=1, . . . , N.sub.UE as follows: $V_{in}\left(m+1\right)={\left[\underset{j=1}{\overset{N_{\mathrm{UE}}}{.Math.}}{\hat{H}}_{\mathrm{jn}}^H{G}_{\mathrm{jn}}^H\left(m\right)G_{\mathrm{jn}}\left(m\right)H_{\mathrm{jn}}+{\upsilon }_{n}I\right]}^{-1}{\hat{H}}_{in}^H{G}_{in}^H\left(m\right)$ c) compute G.sub.in(m+1) using V.sub.in(m+1) and the noise variance estimate σ.sub.i.sup.2 for i=1, . . . , N.sub.UE as follows: $G_{in}\left(m+1\right)=V_{in}^H\left(m+1\right){{\hat{H}}_{in}^H\left[\underset{j=1}{\overset{N_{\mathrm{UE}}}{.Math.}}{\hat{H}}_{in}{V}_{\mathrm{jn}}\left(m+1\right)V_{\mathrm{jn}}^H\left(m+1\right){\hat{H}}_{in}^H+{\sigma }_i^{2}I\right]}^{-1}$ d) compute $E=\underset{i=1}{\overset{N_{\mathrm{UE}}}{.Math.}}{.Math.G_{in}\left(m+1\right)-G_{in}\left(m\right).Math.}_F^2$ e) repeat step b), step c) and step d) until $\underset{i=1}{\overset{N_{\mathrm{UE}}}{.Math.}}{.Math.G_{in}\left(m+1\right)-G_{in}\left(m\right).Math.}_F^2<\mathrm{.Math.},$ where ∥.Math.∥.sub.F.sup.2 denotes the Frobenius norm and ε is a convergent threshold; and f) output V.sub.in(m+1), i=1, . . . , N.sub.UE.

8. The method as claimed in claim 7, wherein the Lagrange multiplier ν.sub.n for the n-th TP is computed using the following procedure: a) compute λ.sub.k as $U\Lambda U^H=\underset{i=1}{\overset{N_{\mathrm{UE}}}{.Math.}}{\hat{H}}_{in}^H{G}_{in}^H\left(m\right)G_{in}\left(m\right){\hat{H}}_{in}$ b) set ν.sub.min and ν.sub.max; c) set ν.sub.n=(ν.sub.max+ν.sub.min)/2; d) compute a quantity ${\hat{P}}_n=\underset{k=1}{\overset{{\tau }_n}{.Math.}}\frac{{\lambda }_k}{{\left({\lambda }_k+{\upsilon }_n\right)}^2};$ e) check if {circumflex over (P)}.sub.n>P.sub.n and if so set ν.sub.min=ν.sub.n otherwise set ν.sub.max=ν.sub.n, where P.sub.n is the transmit power of the n-th TP; f) repeat step c), step d) and step e) until |{circumflex over (P)}.sub.n−P.sub.n|<ε, where ε is a convergent threshold; and g) output ν.sub.n.

9. The method as claimed in claim 4, wherein coherent precoding is used and the method further comprises finding a representative matrix Ĥ.sub.i representing the total channel as follows:
Ĥ.sub.i=[Ĥ.sub.i1,Ĥ.sub.i2, . . . ,Ĥ.sub.iN.sub.TP],i=1, . . . ,N.sub.UE

10. The method as claimed in claim 9, wherein the method further comprises using the representative matrix Ĥ.sub.i, a Lagrange multiplier ν and a noise variance estimate σ.sub.i.sup.2 to compute the precoders (V.sub.i).

11. The method as claimed in claim 10, wherein precoders V.sub.i are computed using an iterative procedure.

12. The method as claimed in claim 11, wherein precoders V.sub.i are computed using the following iterative procedure where (m) denotes the m-th iteration: a) initialize G.sub.i(m=0)=J.sub.i, i=1, . . . , N.sub.UE, where J.sub.i is a RI.sub.i×N.sub.TX matrix with the (a,b)-th element being zero for a≠b and 1 for a=b, and N.sub.TX is the total number of transmit antennas of all TPs; b) compute V.sub.i(m+1) using G.sub.i(m) and the Lagrange multiplier ν for i=1, . . . , N.sub.UE as follows: $V_i\left(m+1\right)={\left[\underset{j=1}{\overset{N_{\mathrm{UE}}}{.Math.}}{\hat{H}}_j^H{G}_j^H\left(m\right)G_j\left(m\right){\hat{H}}_j+\mathrm{\upsilon I}\right]}^{-1}{\tilde{H}}_i^H{G}_i^H\left(m\right)$ c) compute G.sub.i(m+1) using V.sub.i(m+1) and the noise variance estimate σ.sub.i.sup.2 for i=1, . . . , N.sub.UE as follows: $G_i\left(m+1\right)=V_i^H\left(m+1\right){{\hat{H}}_i^H\left[\underset{j=1}{\overset{N_{\mathrm{UE}}}{.Math.}}{\hat{H}}_i{V}_j\left(m+1\right)V_j^H\left(m+1\right){\hat{H}}_i^H+{\sigma }_i^{2}I\right]}^{-1}$ d) compute $E=\underset{i=1}{\overset{N_{\mathrm{UE}}}{.Math.}}{.Math.G_i\left(m+1\right)-G_i\left(m\right).Math.}_F^2$ e) repeat step b), step c) and step d) until $\underset{i=1}{\overset{N_{\mathrm{UE}}}{.Math.}}{.Math.G_i\left(m+1\right)-G_i\left(m\right).Math.}_F^2<\mathrm{.Math.};$ f) output V.sub.i(m+1), i=1, . . . , N.sub.UE.

13. The method as claimed in claim 12, wherein the Lagrange multiplier ν is computed using the following procedure: a) compute λ.sub.k as $U\Lambda U^H=\underset{i=1}{\overset{N_{\mathrm{UE}}}{.Math.}}{\hat{H}}_i^H{G}_i^H\left(m\right)G_i\left(m\right){\hat{H}}_i$ b) set ν.sub.min and ν.sub.max; c) set ν=(ν.sub.max+ν.sub.min)/2; d) compute the following quantity $\hat{P}=\underset{k=1}{\overset{N_{\mathrm{TX}}}{.Math.}}\frac{{\lambda }_k}{{\left({\lambda }_k+\upsilon \right)}^2};$ e) check if {circumflex over (P)}>P then set ν.sub.min=ν otherwise set ν.sub.max=ν, where P is the total transmit power $P=\underset{n=1}{\overset{N_{\mathrm{TP}}}{.Math.}}{P}_n;$ f) repeat step c), step d) and step e) until |{circumflex over (P)}−P|<ε, where ε is a convergent threshold; and g) output the Lagrange multiplier ν.

14. The method as claimed in claim 10, wherein the CSI transmitted from each UE to its serving TP includes a channel quality indicator (CQI) and the noise variance estimate σ.sub.i.sup.2 is found using the CQI as follows: a) find the signal to interference plus noise ratio (SINR.sub.i1) based on the SINR thresholds in the CQI table; and b) calculate σ.sub.i.sup.2 using the SINR.sub.i1 and the serving TP's transmit power P.sub.s as follows: ${\sigma }_i^2=\frac{P_s/N_{\mathrm{UE}}}{\underset{l=1}{\overset{L_i}{.Math.}}{\mathrm{SINR}}_{\mathrm{il}}/L_i},i=1,\mathrm{.Math.},N_{\mathrm{UE}}$ where L.sub.i is the number of codewords used for the i-th UE.

15. The method as claimed in claim 14 wherein, from as many as N.sub.TP reported RI.sub.in, the majority is selected as a single common RI.sub.i for the i-th UE.

16. The method as claimed in claim 15, wherein only CQI.sub.i{circumflex over (n)}(l) associated with the selected RI.sub.i are candidates for CQI selection, the selection is carried out per codeword independently, and the majority among the candidates is selected as the common CQI for the l-th codeword CQI.sub.i(l).

17. The method as claimed in claim 5, wherein the CSI transmitted from each UE to its serving TP includes a channel quality indicator (CQI) and the noise variance estimate σ.sub.i.sup.2 is found using the CQI as follows: a) find the signal to interference plus noise ratio (SINR.sub.i1) based on thresholds in the CQI table; and b) calculate σ.sub.i.sup.2 using the SINR.sub.i1 and the serving TP's transmit power P.sub.s as follows: ${\sigma }_i^2=\frac{P_s/N_{\mathrm{UE}}}{\underset{l=1}{\overset{L_i}{.Math.}}{\mathrm{SINR}}_{\mathrm{il}}/L_i},i=1,\mathrm{.Math.},N_{\mathrm{UE}}$ where L.sub.i is the number of codewords used for the i-th UE.

18. The method as claimed in claim 17 wherein, from up to N.sub.TP reported RI.sub.in, the majority is selected as a single common RI.sub.i for the i-th UE.

19. The method as claimed in claim 18, wherein only CQI.sub.i{circumflex over (n)}(l) associated with the selected RI.sub.i are candidates for CQI selection, the selection is carried out per codeword independently, and the majority among the candidates is selected as a common CQI for the l-th codeword CQI.sub.i(l).

20. A downlink coordinated multi-point transmission/reception (DL CoMP) wireless communications system in which joint transmission (JT) is performed between a plurality of transmission points (TPs) and a plurality of user equipments (UEs), wherein each UE has one of the TPs as its serving TP, channel state information (CSI) is transmitted from each UE to its serving TP, the transmitted CSI includes precoder matrix indicators (PMI), and the PMI is used to generate precoders for transmission of data from the plurality of TPs to the plurality of UEs, wherein using the PMI to generate precoders involves using the PMI to find a representative matrix (Ĥ.sub.in) representing the channel (Ĥ.sub.in) between an n-th TP and an i-th UE, wherein a fixed codebook (Ω.sub.RI) of representative matrices is generated from PMI codebook(s), the CSI transmitted from each UE to its serving TP includes a rank indicator (RI), and Ω.sub.RI is different for different RI, and if the RI for the i-th UE (RI.sub.i) is equal to the number of receive antennas of the UE (N.sub.RX) (i.e. if RI.sub.i=N.sub.RX) then Ω.sub.RI contains matrices Ĥ(m), m=1, . . . of size N.sub.RX×τ.sub.n, where τ.sub.n is the number of antennas at the n-th TP, i=1, . . . , N.sub.UE, n=1, . . . , N.sub.TP where N.sub.TP and N.sub.UE are the number of TPs and UEs, respectively.

Description

BRIEF DESCRIPTION OF DRAWINGS

(1) Preferred features, embodiments and variations of the invention may be discerned from the following Detailed Description which provides sufficient information for those skilled in the art to perform the invention. The Detailed Description is not to be regarded as limiting the scope of the preceding Summary of the Invention in any way. The Detailed Description will make reference to a number of drawings as follows:

(2) FIG. 1 is schematically represents a JT-DL CoMP system.

(3) FIG. 2 is schematically represents the way each UE feeds back CSI to its serving TP via uplink.

(4) FIG. 3A schematically illustrates CSI measurement for JT-DL CoMP.

(5) FIG. 3B schematically illustrates CSI measurement for JT-DL CoMP.

(6) FIG. 4 is a flowchart illustrating, for the case of non-coherent precoding, a method for generating j-MMSE precoders in accordance with the embodiment of the invention discussed below.

(7) FIG. 5 is a flowchart illustrating, for the case of non-coherent precoding, a method for computing the Lagrange multiplier for the n-th TP.

(8) FIG. 6 is a flowchart illustrating, for the case of coherent precoding, a method for generating j-MMSE precoders in accordance with the embodiment of the invention discussed below.

(9) FIG. 7 is a flowchart illustrating, for the case of coherent precoding, a method for computing the Lagrange multiplier for all TPs.

(10) FIG. 8 schematically illustrates the estimation of a UE's noise variance based on the reported CQI of the serving TP.

(11) FIG. 9 schematically illustrates RI and CQI collection from the reported RI and CQI for transmission to the i-th UE.

DESCRIPTION OF EMBODIMENTS

(12) Joint transmit & receive optimisation methods have previously been proposed. See, for example, Sampath H. and Paulraj A., “Joint Transmit and Receive Optimization for High Data Rate Wireless Communication Using Multiple Antennas”, Thirty-Third Asilomar Conference on Signals, Systems, and Computers, 1999, and Zhang J., et. al., “Joint Linear Transmitter and Receiver Design for Downlink of Multiuser MIMO Systems”, IEEE Communications Letters, Vol. 9, No. 11, November 2005.

(13) Embodiments of the present invention provide MMSE precoders based (at least somewhat) on the joint transmit & receive optimization methods discussed in the above academic papers. However, unlike the methods in these academic papers, the present invention does not require knowledge of the channel to generate MMSE precoders. Instead (and in contrast), embodiments of the invention require only the PMI, which is fed back by UEs to serving TPs, as shown in FIG. 2. The precoder according to the particular embodiments of the invention discussed below will be referred to as the j-MMSE precoder.

(14) A) Non-Coherent Precoding

(15) In the case of non-coherent precoding, the individual j-MMSE precoder V.sub.in is computed using the joint transmit and receive MMSE optimization as follows.

(16) Finding Representative Channels

(17) Let Ω.sub.RI denote the fixed codebook of representative channel matrices which is generated from the PMI codebook(s). There are different Ω.sub.RI for different RI. For RI.sub.i=N.sub.RX, the Ω.sub.RI contains matrices Ĥ(m), m=1, . . . of size N.sub.RX×τ.sub.n For RI.sub.i<N.sub.RX, the Ω.sub.RI contains vectors ĥ(m), m=1, . . . of size τ.sub.n×1

(18) Let Ĥ.sub.in be the representative for the channel H.sub.in. The representative channel is obtained as follows:

(19) If RI.sub.i=N.sub.RX, then
Ĥ.sub.in=Ĥ(m*)εΩ.sub.RI,i=1, . . . ,N.sub.UE,n=1, . . . ,N.sub.TP
with

(20) $\begin{array}{cc}{m}^{*}=\underset{m}{\arg \max }\mathrm{tr}\left\{{\left[\hat{H}\left(m\right)W_{\mathrm{in}}\right]}^H\left[\hat{H}\left(m\right)W_{\mathrm{in}}\right]\right\},\hat{H}\left(m\right)\in {\Omega }_{\mathrm{RI}}& \mathrm{Eqaution}\left(4\right)\end{array}$

(21) If RI.sub.i<N.sub.RX, then

(22) 1) Calculate correlation values:
C.sub.in(m)=tr{[ĥ.sup.H(m)W.sub.in].sup.H[ĥ.sup.H(m)W.sub.in]},m=1,  Equation (5)
and

(23) 2) Sort to find the N.sub.RX correlation values C.sub.in(m.sub.1)>C.sub.in(m.sub.2)> . . . >C.sub.in(m.sub.N.sub.RX) and the N.sub.RX corresponding ĥ(m.sub.1), ĥ(m.sub.2), . . . , ĥ(m.sub.N.sub.RX) to form the channel matrix
Ĥ.sub.in=[ĥ(m.sub.1),ĥ(m.sub.2), . . . ,ĥ (m.sub.N.sub.RX)].sup.H  Equation (6)

(24) Here W.sub.in (of size τ.sub.n×RI.sub.i) is the precoder in the 3GPP standard (TS 36.211) associated with the PMI p.sub.in. Note that, if the PMI consists of PMI#1 and PMI#2, then W.sub.in=W.sub.in(1)×W.sub.in(2).

(25) Generating the j-MMSE Precoder V.sub.in (see FIG. 4)

(26) Let (m) denote the m-th iteration of the procedure. The precoder is generated as follows: 1) (401) Initialize G.sub.in (m=0)=J.sub.in, i=1, . . . , N.sub.UE. Here J.sub.in is a RI.sub.i×τ.sub.n matrix with the (a,b)-th element being zero for a≠b and being 1 for a=b. 2) (402) Compute V.sub.in(m+1) using G.sub.in(m) and the Lagrange multiplier ν.sub.n for i=1, . . . , N.sub.UE as follows.

(27) $\begin{array}{cc}{V}_{in}\left(m+1\right)={\left[\underset{j=1}{\overset{N_{\mathrm{UE}}}{.Math.}}{\hat{H}}_{\mathrm{jn}}^H{G}_{\mathrm{jn}}^H\left(m\right)G_{\mathrm{jn}}\left(m\right){\hat{H}}_{\mathrm{jn}}+{\upsilon }_{n}I\right]}^{-1}{\hat{H}}_{in}^H{G}_{in}^H\left(m\right)& \mathrm{Equation}\left(7\right)\end{array}$ 3) (403) Compute G.sub.in(m+1) using V.sub.in(m+1) and the given noise variance estimate σ.sub.i.sup.2 for i=1, . . . , N.sub.UE as follows.

(28) $\begin{array}{cc}{G}_{in}\left(m+1\right)=V_{in}^H\left(m+1\right){{\hat{H}}_{in}^H\left[\underset{j=1}{\overset{N_{\mathrm{UE}}}{.Math.}}{\hat{H}}_{in}{V}_{\mathrm{jn}}\left(m+1\right)V_{\mathrm{jn}}^H\left(m+1\right){\hat{H}}_{in}^H+{\sigma }_i^{2}I\right]}^{-1}& \mathrm{Equation}\left(8\right)\end{array}$ 4) (404) Compute

(29) $E=\underset{i=1}{\overset{N_{\mathrm{UE}}}{.Math.}}{.Math.G_{in}\left(m+1\right)-G_{in}\left(m\right).Math.}_F^2$ 5) (405) increment m and repeat step 2), step 3) and step 4) until

(30) $\underset{i=1}{\overset{N_{\mathrm{UE}}}{.Math.}}{.Math.G_{in}\left(m+1\right)-G_{in}\left(m\right).Math.}_F^2<\mathrm{.Math.}.$
Here ∥.Math.∥.sup.2.sub.F denotes Frobenius norm and ε is the convergent threshold. 6) (406) Output V.sub.in(m+1), i=1, . . . N.sub.UE.

(31) Computing the Lagrange multiplier ν.sub.n (see FIG. 5)

(32) For each of the n-th TP, the Lagrange multiplier ν.sub.n is computed as follows. 1) (501) Compute λ.sub.k as

(33) $\begin{array}{cc}U\Lambda U^H=\underset{i=1}{\overset{N_{\mathrm{UE}}}{.Math.}}{\hat{H}}_{in}^H{G}_{in}^H\left(m\right)G_{in}\left(m\right){\hat{H}}_{in}& \mathrm{Equation}\left(9\right)\end{array}$ 2) (502) Set ν.sub.min and ν.sub.max 3) (503) Set ν.sub.n=(ν.sub.max+ν.sub.min)/2. 4) (504) Compute the following quantity

(34) 0${\hat{P}}_n=\underset{k=1}{\overset{{\tau }_n}{.Math.}}\frac{{\lambda }_k}{{\left({\lambda }_k+{\upsilon }_n\right)}^2}.$ 5) (505) Check if {circumflex over (P)}.sub.n>P.sub.n and if so set ν.sub.min=ν.sub.n otherwise set ν.sub.max=ν.sub.n. Here P.sub.n is the transmit power of the n-th TP. 6) (506) Repeat step 3), step 4) and step 5) until |{circumflex over (P)}.sub.n−P.sub.n|<ε. Here ε is the convergent threshold. 7) (507) Output ν.sub.n.
B) Coherent Precoding

(35) In the case of coherent precoding, the total j-MMSE precoder V, is computed using the joint transmit and receive MMSE optimization as follows:

(36) Finding Representative Channels

(37) First the individual representative channel Ĥ.sub.in is found as in the non-coherent case discussed above. Then the total channel is generated by:
Ĥ.sub.i=└Ĥ.sub.i1,Ĥ.sub.i2, . . . ,Ĥ.sub.iN.sub.TP┘,i=1, . . . ,N.sub.UE  Equation (10)

(38) Generating the j-MMSE Precoder V.sub.i (see FIG. 6)

(39) Let (m) denote the m-th iteration of the procedure. The precoder is generated as follows: a) (601) Initialize G.sub.i(m=0)=J.sub.i, i=1, . . . , N.sub.UE. Here J.sub.i is a RI.sub.i×N.sub.TX matrix with the (a,b)-th element being zero for a≠b and being 1 for a=b. b) (602) Compute V.sub.i(m+1) using G.sub.i(m) and the Lagrange multiplier ν for i=1, . . . , N.sub.UE as follows.

(40) $\begin{array}{cc}{V}_i\left(m+1\right)={\left[\underset{j=1}{\overset{N_{\mathrm{UE}}}{.Math.}}{\hat{H}}_j^H{G}_j^H\left(m\right)G_j\left(m\right){\hat{H}}_j+\upsilon I\right]}^{-1}{\hat{H}}_i^H{G}_i^H\left(m\right)& \mathrm{Equation}\left(11\right)\end{array}$ c) (603) Compute G.sub.i(m+1) using V.sub.i(m+1) and the given noise variance estimate σ.sub.i.sup.2 for i=1, . . . , N.sub.UE as follows.

(41) $\begin{array}{cc}{G}_i\left(m+1\right)=V_i^H\left(m+1\right){{\hat{H}}_i^H\left[\underset{j=1}{\overset{N_{\mathrm{UE}}}{.Math.}}{\hat{H}}_i{V}_j\left(m+1\right)V_j^H\left(m+1\right){\hat{H}}_i^H+{\sigma }_i^{2}I\right]}^{-1}& \mathrm{Equation}\left(12\right)\end{array}$ d) (604) Compute

(42) $E=\underset{i=1}{\overset{N_{\mathrm{UE}}}{.Math.}}{.Math.G_i\left(m+1\right)-G_i\left(m\right).Math.}_F^2$ e) (605) increment m and repeat step b), step c) and step d) until

(43) $\underset{i=1}{\overset{N_{\mathrm{UE}}}{.Math.}}{.Math.G_i\left(m+1\right)-G_i\left(m\right).Math.}_F^2<\mathrm{.Math.}.$ f) (606) Output V.sub.i(m+1), i−1, . . . N.sub.UE

(44) Computing the Lagrange Multiplier ν (see FIG. 7)

(45) The Lagrange multiplier ν is obtained as follows. 1) (701) Compute λ.sub.k as

(46) $\begin{array}{cc}U\Lambda U^H=\underset{i=1}{\overset{N_{\mathrm{UE}}}{.Math.}}{\hat{H}}_i^H{G}_i^H\left(m\right)G_i\left(m\right){\hat{H}}_i& \mathrm{Equation}\left(13\right)\end{array}$ 2) (702) Set ν.sub.min and ν.sub.max 3) (703) Set ν=(ν.sub.max+ν.sub.min)/2. 4) (704) Compute the following quantity

(47) $\hat{P}=\underset{k=1}{\overset{N_{\mathrm{TX}}}{.Math.}}\frac{{\lambda }_k}{{\left({\lambda }_k+\upsilon \right)}^2}.$ 5) (705) Check if {circumflex over (P)}>P and if so set ν.sub.min=ν otherwise set ν.sub.max=ν. Here P is the total transmit power,

(48) $P=\underset{n=1}{\overset{N_{\mathrm{TP}}}{.Math.}}{P}_n.$ 6) (706) Repeat step 3), step 4) and step 5) until |{circumflex over (P)}−P|<ε. Here ε is the convergent threshold. 7) (707) Output the Lagrange multiplier ν.
C) Noise Variance Estimating (see FIG. 8)

(49) The following noise variance estimation may be used for both non-coherent and coherent precoding. The method estimates the UE's noise variance from the reported CQI for the serving TP is as follows: 1) (801) Find SINR.sub.i1 based on the SINR thresholds in the CQI table. 2) (802) Calculate σ.sub.i.sup.2 using SINR.sub.i1 and the serving TP's transmit power P.sub.s as follows.

(50) $\begin{array}{cc}{\sigma }_i^2=\frac{P_s/N_{\mathrm{UE}}}{\underset{l=1}{\overset{L_i}{.Math.}}{\mathrm{SINR}}_{\mathrm{il}}/L_i},i=1,\mathrm{.Math.},N_{\mathrm{UE}}& \mathrm{Equation}\left(14\right)\end{array}$
where L.sub.i the number of codewords used for the i-th UE. For less complexity, the noise variance can be fixed to zero as:
σ.sub.i.sup.2=0,i=1, . . . ,N.sub.UE  Equation (15)
D) Rank and CQI selection (see FIG. 9)

(51) Because, for a given UE, all TPs have common transmission rank and common CQI, it follows that rank and CQI selection is necessary. From the as many as N.sub.TP reported RI.sub.in, the majority is selected as the single common RI.sub.i for the i-th UE. The selection can be done using the histogram. Then only CQI.sub.i{circumflex over (n)} (l) associated with the selected RI.sub.i are the candidates for CQI selection. The selection is carried out per codeword independently. The majority among the candidates is selected as the common CQI for the l-th codeword CQI.sub.i(l). The selection can be done using the histogram.

(52) Advantages

(53) As discussed above, embodiments of the present invention do not require knowledge of the channel to generate the j-MMSE precoder. Rather, they require only the PMI which is fed back by UEs. This may provide a number of advantages. For instance, it may provide improved performance in comparison with methods which directly use reported PMI. Also, as is made evident above, the invention is applicable to both coherent and non-coherent precoding.

(54) In the present specification and claims (if any), the word ‘comprising’ and its derivatives including ‘comprises’ and ‘comprise’ include each of the stated integers but does not exclude the inclusion of one or more further integers.

(55) Reference throughout this specification to ‘one embodiment’ or ‘an embodiment’ means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. Thus, the appearance of the phrases ‘in one embodiment’ or ‘in an embodiment’ in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more combinations.

(56) In compliance with the statute, the invention has been described in language more or less specific to structural, systems or methodical features. It is to be understood that the invention is not limited to specific features shown or described since the means herein described comprises preferred forms of putting the invention into effect. The invention is, therefore, claimed in any of its forms or modifications within the proper scope of the appended claims (if any) appropriately interpreted by those skilled in the art.

(57) This application is based upon and claims the benefit of priority from Australia Patent Application No. 2013902955, filed on Aug. 7, 2013, the disclosure of which is incorporated herein in its entirety by reference.

REFERENCE SIGNS LIST

(58) TP0 SERVING TP TP1 NEIGHBOURING TP